Advances in Information and Communication Technology and Systems [1st ed.] 9783030583583, 9783030583590

Citation preview

Lecture Notes in Networks and Systems 152

Mykhailo Ilchenko Leonid Uryvsky Larysa Globa   Editors

Advances in Information and Communication Technology and Systems

Lecture Notes in Networks and Systems Volume 152

Series Editor Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland Advisory Editors Fernando Gomide, Department of Computer Engineering and Automation—DCA, School of Electrical and Computer Engineering—FEEC, University of Campinas— UNICAMP, São Paulo, Brazil Okyay Kaynak, Department of Electrical and Electronic Engineering, Bogazici University, Istanbul, Turkey Derong Liu, Department of Electrical and Computer Engineering, University of Illinois at Chicago, Chicago, USA, Institute of Automation, Chinese Academy of Sciences, Beijing, China Witold Pedrycz, Department of Electrical and Computer Engineering, University of Alberta, Alberta, Canada, Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland Marios M. Polycarpou, Department of Electrical and Computer Engineering, KIOS Research Center for Intelligent Systems and Networks, University of Cyprus, Nicosia, Cyprus Imre J. Rudas, Óbuda University, Budapest, Hungary Jun Wang, Department of Computer Science, City University of Hong Kong, Kowloon, Hong Kong

The series “Lecture Notes in Networks and Systems” publishes the latest developments in Networks and Systems—quickly, informally and with high quality. Original research reported in proceedings and post-proceedings represents the core of LNNS. Volumes published in LNNS embrace all aspects and subfields of, as well as new challenges in, Networks and Systems. The series contains proceedings and edited volumes in systems and networks, spanning the areas of Cyber-Physical Systems, Autonomous Systems, Sensor Networks, Control Systems, Energy Systems, Automotive Systems, Biological Systems, Vehicular Networking and Connected Vehicles, Aerospace Systems, Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power Systems, Robotics, Social Systems, Economic Systems and other. Of particular value to both the contributors and the readership are the short publication timeframe and the world-wide distribution and exposure which enable both a wide and rapid dissemination of research output. The series covers the theory, applications, and perspectives on the state of the art and future developments relevant to systems and networks, decision making, control, complex processes and related areas, as embedded in the fields of interdisciplinary and applied sciences, engineering, computer science, physics, economics, social, and life sciences, as well as the paradigms and methodologies behind them. ** Indexing: The books of this series are submitted to ISI Proceedings, SCOPUS, Google Scholar and Springerlink **

More information about this series at http://www.springer.com/series/15179

Mykhailo Ilchenko Leonid Uryvsky Larysa Globa •

•

Editors

Advances in Information and Communication Technology and Systems

123

Editors Mykhailo Ilchenko Igor Sykorsky Kyiv Polytechnic Institute National Technical University of Ukraine Kyiv, Ukraine

Leonid Uryvsky Igor Sykorsky Kyiv Polytechnic Institute National Technical University of Ukraine Kyiv, Ukraine

Larysa Globa Igor Sykorsky Kyiv Polytechnic Institute National Technical University of Ukraine Kyiv, Ukraine

ISSN 2367-3370 ISSN 2367-3389 (electronic) Lecture Notes in Networks and Systems ISBN 978-3-030-58358-3 ISBN 978-3-030-58359-0 (eBook) https://doi.org/10.1007/978-3-030-58359-0 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

This volume is a collection of the most important research results in fields of Information, Telecommunication and RadioElectronics Technologies provided by different group of researchers from Ukraine in collaboration with scientists from different countries. The authors of the chapters from this collection present in-depth extended research results in their scientific fields. The volume consists of three parts. Part I Modern Challenges in Information Technologies deals with various aspects to the analysis and solution of practically important issues of information systems in general, contains an overview of global trends for the development of information and communication technologies, as well as progression from big data to smart data, development of cloud-based architecture, practical implementation of Internet of things (IoT) cryptanalysis, biometric cryptosystems, realization of biometric authentication, fundamentals of information and analytical activities, including scientific and educational portals, in particular. Part II Modern Challenges in Telecommunication Technologies contains original works dealing with many aspects of construction, using research and forecasting of technological and services characteristics of telecommunication systems. The presented studies of this part cover a wide range of telecommunication technologies, including multigigabit wide area networks, wireless communication systems, multiservice transmission systems, as well as in-depth aspects of the study of methods in problems of forecasting, the implementation of biometric data security in telecommunications. Part III Modern Challenges in RadioElectronics Technologies contains actual papers, which show some effective technological solutions that can be used for the implementation of novel systems. For the convenience of the readers, we briefly summarize contents of the chapters accepted.

v

vi

Preface

Part I Modern Challenges in Information Technologies. The first chapter presented by M. Ilchenko, L. Uryvsky, S. Osypchuk The Main Directions of Improving Information and Communication Technologies in the Global Trends introduces an analysis of modern and promising directions of formation, processing and transmission of information. Information and telecommunication technologies (ITT) in the light of global trends are considered as a single technological complex. Key trends in the development of ITT identified, including, in the light of the new breakthrough achievements of 2019, which are associated with the further development of ITT. Particular accent is attach to the development of mobile telecommunications (technology 5G and 6G), the development of IoT, the penetration of ITT into the development of Industry 4.0, and the integration of the global satellite Internet. The active desire of Ukraine in the structure of ITT to be at the level of world modern trends is emphasized. The chapter From Big Data to Smart Data: the most effective Approaches for Data Analytics by A. Luntovskyy, L. Globa, B. Shubyn focuses on the problem of thick and server-less mobile applications (Cyber-PHY, IoT, sensor networks, robotics), real-time network applications (thin clouds clients), that can generate large arrays of unmanaged, weakly structured, and non-configured data of various types, known as “big data.” With the acceleration of industrial development “Industry 4.0” processing of such data became considerably more complicated. However, the so-called problem “big data” is hard to solve or resist nowadays. The paper discusses the best practises and case studies for data analytics aimed to overcoming of the big data problematics under a slogan: “From Big Data to Smart Data!” The chapter Cloud-Based Architecture Development to share Vehicle and Traffic Information for Industry 4.0 by Ibrahim Serhat Bulut, Haci Ilhan covers the automotive and technology sectors that have developed very rapidly over the past decade. In addition to this growth, it also introduced the term Industry 4.0, which is used to represent the current Industrial Revolution. This revolution encompasses many sectors from manufacturing to health care. With Industry 4.0, digital transformation can create value throughout the entire product lifecycle, support customer feedback, and provide advanced solutions to the problems to be experienced. Automatic communication between the vehicle and the management office will facilitate our lives by enabling different analysis of vehicles such as the driver’s vehicle usage history, fuel consumption, maintenance indicators, and determination of a behavioral model with data like temperature. This provides analysis of the car, the driver’s experience, and the road, preventing critical problems and unwanted behavior, and increasing safety on the roads. For example, during the winter season, municipal employees will not have to wait at night to intervene in road freezing. Employees will instantly monitor which roads are at risk of icing through the data collected from the vehicles on the road and municipal workers will salinize them in no time. This article aims to implement a platform that collects and analyzes vehicle sensor data and provides individual and corporate feedback. Using the OBD-II scanner, it is intended to help prevent problems, reduce

Preface

vii

accident rates, and manage different types of vehicles. A ready OBD-II (On Board Diagnostic) reader device, sup-ported by a Bluetooth connection, is used to collect data directly from the Engine Control Unit (ECU) in real-time, using an androidbased smartphone as a cloud network connection. The architectural structure in the cloud is capable of collecting and analyzing raw data to detect the occurrence of errors in vehicles. While providing feedback to the user, a smart cloud-based architecture provides the necessary information to the relevant municipal, fire or ambulance units by foreseeing traffic accidents, the icing on roads in winter or asphalt melting in summer. Experiments and tests conducted in Istanbul’s traffic show that the proposed platform has applicability and potential to use. In the chapter IoT Cryptanalysis using Neural Cryptography and PostQuantum Cryptography by L. Mirtskhulava, N. Gulua, L. Globa, N. Meshveliani, the authors investigated the features IoT as a powerful tool to capture data and analyze them from connected devices capable optimizing the processes that improves customer service but exposes to security vulnerabilities. IoT spans personal and sensitive information of the users and many aspects of private life of the persons and therefore IoT needs to be protect. Assessing IoT security, we can evaluate the ecosystems of IoT among billions of the devices connected through the Internet prominent in security analysis of the devices. In IoT security, both the network security and the devices security are so vital. Analysis of IoT security using probabilistic NTRU cryptosystem and using artificial neural network algorithms in cryptanalysis and data encryption is introduce. The authors present two approaches that can withstand communication attacks. Both cryptography technologies like neural cryptography and post-quantum cryptography are the best solutions we’re using in the given paper. Using NTRU, we implement the NTRUEncrypt public key encryption algorithm capable of standing against the attacks from the quantum computers. NTRU allows longterm cryptanalysis and is significantly faster than the other public-key cryptosystems such as RSA or ECC. The other hand, it is well known that artificial neural networks are able to explore the solution in the field of cryptanalysis capable to attack ciphering algorithms where any of the functions can be produced by a neural network. Therefore, learning ability of neural networks is used in public-key cryptography. In result, using NTRU, it made possible to change keys in a few seconds that making difficult for an attacker to crack the encryption scheme. Collected dataset was used to train neural networks. ML classifier has achieved an accuracy of 99.9%. Unsupervised DL algorithms were proposed for detecting attacks in IoT. NTRU was offered by evaluation of other asymmetric algorithms such as RSA, ECC. Proposed cryptosystems are the better solution in encryption, decryption, and key generation. NTRU does not suffer with factorization. So, asymmetric key protocol with light weight as NTRU is better solution for securing IoT network. In the other hand, neural cryptography gave an opportunity to use neural key exchange protocol that makes faster key exchange schemes. Unsupervised deep learning algorithms can faster detect attacks in IoT networks.

viii

Preface

In Biometric Cryptosystems: Overview, State-of-the-Art and Perspective Directions by M. Lutsenko, A. Kuznetsov, A. Kiian, O. Smirnov, T. Kuznetsova, the authors present modern cryptographic systems that are constantly evolving and improving. This is due to the development of new computing systems and advanced cryptographic analysis methods, as well as increasing requirements for speed, security and reliability of used tools. In particular, it was announced an advent of universal quantum computers, which will be able to provide cryptanalysis with advanced calculation methods based on fundamentally new physical principles. The possible use of such devices encourages the development, research, and standardization of algorithms for post-quantum cryptographic information protection. Another factor for the development of advanced cryptographic systems is the biometric technologies popularization. In this work, a critical review and analysis of the current application of biometric technologies in cryptographic systems is conducted. In particular, biometric cryptographic systems, which are designed to generate secure pseudorandom sequences that can be used as cryptographic keys, passwords etc., are investigated. A comparative analysis of various biometric cryptosystems with the determination of their advantages and disadvantages is carried out. The perspective directions for further research are substantiated. Also, this work presents a new key generation scheme which uses fuzzy extractors from the biometric data of iris. The proposed method is based on the code-based public key cryptosystems which are considered to be resistant to quantum cryptanalysis. A software implementation of this method with experimental studies of the key generation algorithm and recommendations on the practical application are proposed. A. Kuznetsov, I. Oleshko, K. Chernov, M. Bagmut, T. Smirnova (Biometric Authentication Using Convolutional Neural Networks) discuss an approach to analysis of biometric identity authentication technologies that are widespread. These systems are implemented not only in enterprises, controlled-access facilities, but also on smartphones of ordinary users and in online applications. The problem of choosing one of the authentication methods remains urgent. This paper provides a comparative analysis of existing systems and concludes that one of the most common and persistent methods is facial authentication system. The most powerful types of attacks on the biometric system are attacks on the database of biometric templates and attacks on sensors for obtaining biometric characteristics. Attacks on biometric sensors or spoofing attack are aimed at impersonating another person through fake biometric data. The paper deals with the possibility of special attacks on the biometric system of authentication by face image. A new method of detecting fake attacks (spoofing attacks) is proposed. The method is based on the use of an artificial convolutional neural network, which was trained using a ReplayAttack Database from Idiap Research Institute. The obtained results show high efficiency of the proposed method of detecting spoofing attacks: the probability that an attack will be detected is 94.98%. The chapter Transdisciplinary Fundamentals of Information-Analytical Activity by S. Dovgyi, O. Stryzhak focuses on the problem of accumulation of large volumes of scientific and technical products (STP), which is a passive

Preface

ix

distributed system of knowledge. This research considers transdisciplinary of such information resources as a metacategory that takes into account the hyperproperties of big data (big data), namely: a) reflection which implements the principles of integration, the consistency and the their behavior integrity and guarantee; b) recursion which implements the recurrence category of their operational transformation forms during activation; c) reduction on the basis of which these forms decomposition principle is realized. Their interpretation in the case of big data processing is implemented in the following areas: i) the information resources structural analysis; ii) forms of interaction with information resources; iii) definition of the mechanisms for identifying criteria for selecting appropriate contexts that needed for the expert analysis. The actuality of the such procedures implementation is based on the need to create the conditions for supporting the effective a large number of diverse information arrays processing for the information and analytical activities of experts. This understanding for solving the Big Data processing problem is supported by implementation of component architecture of the services to support the analytical processes of the experts from various thematic sphere of activity. In Ontological Fundamentals of Scientific and Education Portals by S. Dovgyi, N. Gayevska, O. Lisovyi, Iu. Mosenkis, the authors present the methodology and technological solutions for design and creation of web-based scientific and educational portals of ontological engineering. Web portals are defined as network systems of aggregated representation of subject systems of knowledge. A narrative category is entered, as an integrated representation of descriptions of passive knowledge systems. All stages of implementation are considered—the analysis of subject areas, by content of which creates all the portal system components, design the structure and functional on the basis of ontological modeling of subject areas, highlighting cross-curricular relationships to determine conditions of transition in the portal information environment from one state to another. Mechanisms of dynamic cataloging of information resources of scientificeducational portal, cluster formation by thematic search profiles for relevant information are presented. The model of aggregate displaying portal thematic resources is given. Formation means in the use of scientific and educational portals on the base of the component architecture of WEB-semantics services are determined and it describes their mechanisms application with the support of research and educational-cognitive activity of portal users. Based on the variety of descriptions of creative activity of Taras Shevchenko, examples of reflection his legacy in the information environment of the scientific-educational portal are given. Ontological means of dialogue with the image of the writer on the basis of establishing intertextual links with his works are presented. Part II Modern Challenges in Telecommunication Technologies. K. Karpov, D. Syzov, V. Kirova, N. Mareev, E. Siemens (Data Transmission Performance Enhancement in Multi-gigabit Wide Area Networks) present the scientific and technical principles of growing information volumes while meeting the requirements for their quality. Over the past few years, amount of produced data

x

Preface

and WAN users experienced rapid growth, e.g., according to IBM Cloud Marketing research amount of Internet Users was grown from 2.4 billion in 2014 to 4.4. billion by 2019. Over 90% of worldwide data was generated since 2016 with the every-day replenishment of 2.5 quintillion bytes. Consequently, the demand for technologies capable of handling such amounts of traffic increases as well. At the same time, the requirements for quality provided by those technologies changed. Partially due to changes in types of Internet traffic, for example, a larger part of modern networks is filled with media traffic; some new technologies like IoT require low-latency transmission. Another part of the issue is efficiency as with the increase in data rates, overall losses from inefficient architecture increase as well, which may lead to financial losses. While these issues could be solved with specialized hardware, such an approach is less flexible and requires massive financial investments. Thus, many service providers want a solution that would utilize a provider’s existing hardware infrastructure. This article presents an overview of software solutions developed for handling these issues without significant changes in existing architecture. The complex of solutions aims at different parts of the general problem and consists of algorithms that handle network congestion, acceleration of packet processing, efficient transport protocol, methods of traffic distribution and network analysis. In chapter System of Solutions the Maximum Number of Disjoint Paths Computation under Quality of Service and Security Parameters by O. Lemeshko, O. Yeremenko, M. Yevdokymenko, B. Sleiman, the authors propose a system of solutions to the maximum number of disjoint paths computation under the quality of service and security parameters. The field of application of the calculation models the set of disjoint paths has been explained during the provision of the network capabilities such as quality of service (QoS) and network security by the means of the OSI network layer, namely routing and traffic management protocols and technological solutions. Here, secure routing means aimed at improving network security in terms of the probability of compromising the transmitted confidential data. Toward improving network performance and providing the demanded level of QoS in the network, the QoS-based multipath routing with support of load balancing and traffic engineering concept is used. The basic mathematical model for calculating the maximum number of disjoint paths has been presented. The tasks stated were reduced to solving the optimization problems of integer and mixed integer linear programming with maximization of the number of paths and their bandwidth and minimization of their compromising probability in the presence of linear constraints since the routing variables are Boolean, and variables that determine the number of routes used take only integer values. The numerical examples demonstrate the adequacy of the proposed system of solutions in terms of the correctness of the obtained calculation results. L. Globa, M. Skulysh, E. Siemens (Conditionally infinite telecommunication resource for subscribers) discuss approach to the analysis of factors providing for users of modern communication systems of sufficient telecommunication resource. To ensure guaranteed quality, services are combined into groups. A group of similar services with the same service process requirements is called a slice. Using software-defined networks (SDN) allows to deploy a management system slices and

Preface

xi

telecommunications resources that are allocated for their maintenance. This is possible only using the latest telecommunication technologies, and seamless user connection management. The article proposes an original methodology for managing the servicing streams process in network service providers tunnels. The methodology uses the allowable service node load calculation, overload prediction and live migration algorithms to seamlessly change the service node for the flow. In chapter Different extrapolation methods in Problems of Forecasting by I. Strelkovskaya, I. Solovskaya, A. Makoganiuk considered the task of forecasting characteristics of self-similar traffic in IoT network objects with a significant number of pulsations and the property of long-term dependence, which makes it difficult to forecast in practice. Using the different extrapolation method of splineextrapolation based on linear, cubic and B-cubic spline function and waveletextrapolation based on Haar-wavelet, the results of forecasting of self-similar traffic are obtained. The comparison made allowed the results of traffic forecasting based on the Haar-wavelet and the linear, cubic and B-cubic spline-function using wavelet- and spline-extrapolation. This will allow you to choose one or another extrapolation method to improve the accuracy of the forecast, while ensuring scalability and the ability to use it for various IoT applications to prevent network overloads. In chapter Methods for calculating the performance indicators of IP Multimedia Subsystem (IMS) by Romanov O., Nesterenko M., Veres M., Kamarali R., Saichenko L. considered some aspects related to the problems of convergence of networks of various technologies, ensuring their compatibility in management, signaling and data traffic, providing users with modern services with specified QoS indicators. In the process of implementing 4G and 4.5G technologies in the networks of mobile operators, it turned out that the voice traffic service is possible only when using a core network with switching channels of 2G and 3G technologies. And in order for voice traffic to be serviced in batch mode, it is necessary to bring the network of the mobile operator in accordance with the requirements of the IMS architecture. The implementation of IMS provides a solution to two main tasks. The first is the solution to the problem of interaction between networks of various technologies at the level of signal flows. The second task is to ensure the interaction of networks of various technologies at the level of data flows. These tasks are solved at the control level, the main functional elements of which are the functional blocks CSCF, P-CSCF, I-CSCF, S-CSCF. When designing the IMS management level, it is necessary to determine the amount of traffic that will be served by these CSCF functional elements. Then, it is necessary to calculate the required performance of the elements for serving signal traffic and data streams. At the same time, it is very important to provide the required reliability indicators for the IMS management level. The article analyzes the functioning of the IMS core in solving problems of combining signaling traffic and data networks of various technologies. The list of the main elements involved in the process of solving management problems is determined. Models are proposed that allow calculating the amount of incoming traffic from heterogeneous networks and the amount of traffic after it is converted to the standard form of IMS architecture. To

xii

Preface

ensure the indicated reliability indicators of IMS, the use of methods of reservation of functional elements of the control level is provided. Models have been developed and analytical expressions have been obtained for calculating the values of reliability indicators using various reservation methods. In chapter Implementation Biometric Data Security in Remote Authentication Systems via Network Steganography by G. Liashenko, A. Astrakhantsev, the authors are concerned with some issues related to the need of protection user biometric information increases in consequence of the development of biometric authentication systems. This is due to the fact that fingerprints, iris patterns, face geometry and other biometric data are unique and cannot be replaced. There are various methods for protecting biometric data, but the probability of compromise remains when this data is transmitted over the network. The article presents a method of remote biometric authentication using network steganography for different systems. This improves the reliability of the protection of user biometric data. Analysis of existing methods of network steganography, methods of biometric authentication. Synthesis of new method of remote authentication that will increase the security of user biometric data from unauthorized access. Modeling a remote authentication system using network steganography methods and user biometric data is the method of this work. The common methods of biometric authentication, existing methods for their protection and existing network steganography methods were analyzed. The method of remote biometric authentication using network steganography for various systems is presented. The remote authentication system using network steganography was simulated. The resistance of the investigated methods to detection was evaluated. The effectiveness of the proposed method for protecting biometric data was investigated. L. Afanasieva and S. Kravchuk (Wireless Systems with New Cooperative Relaying Algorithm) present a novel approach for the problem of using a mobile terminal as a relay node allows another mobile terminal to transmit or receive data on two independent communication lines, realizing additional diversity of signals both in space and in time. This signal transmission procedure is known as cooperative relay technology. 3GPP group, whose applying is being researched and developed in a number of scientific papers, incorporates cooperative relaying technology in the standard 5G. This technology is used to combat signal degradation on radio link, especially in a multi-user network environment, allows to reduce the effect of interference at the edges of the cell, and to improve radio transmission parameters, thereby ensuring the necessary quality of service QoS. Due to growing the number of wireless network devices in transmission area between the sender and the target receiver, there are more potential relay nodes. Therefore, the task of choosing the one for implementing cooperative transmission in practice becomes an important task. The paper presents a new method for selecting the best relay node, which takes into account a number of criteria, namely: node position, bit error rate, energy consumption. In chapter Autonomous Unmanned Aerial Vehicles Communications on the Base of Software-Defined Radio by M. Kaydenko and S. Kravchuk, the authors are concerned with some issues related to the current state and new principles of the

Preface

xiii

functional and structural construction of a communication part of the UAV system. This system based on SDR and SoC technologies while ensuring the survivability of the formed radio channels. The main increase in survivability is achieved through the use of adaptation over the frequency range using an additional reception channel for continuous analysis of interference conditions, using two or more data transmission channels, complex algorithms for optimal selection of the operating range, transmission channel parameters and adaptive protocols for simultaneous data transmission over two or more channels. O. Tsukanov, E. Yakornov (Estimation of the motion parameters of the UAVs FANET using the dynamic filtering method) present the scientific and technical principles and the algorithm for estimating the motion parameters of unmanned aerial vehicles as elements of flying wireless sensor networks FANET. To estimate the motion parameters, the proposed stable dynamic filtering algorithm allows us to obtain estimates of coordinates and their derivatives. The results obtained can simultaneously improve the accuracy of the estimation of motion parameters, ensure the sustainability of the assessment process and the efficiency of managing the elements of the wireless sensor network FANET. Part III Modern Challenges in RadioElectronics Technologies. The chapter Universal Complex Model for Estimation the Beam Current Density of High Voltage Glow Discharge Electron Guns by I. Melnyk, S. Tyhai, A. Pochynok covers the universal complex model to estimation the focal parameters of electron beams, formed by high voltage glow discharge electron guns. This complex model consists on the means for calculation of electric field distribution and charged particles trajectories with taking into account the space charge, means for defining the current density distribution in the beam focus on the outlet of electron gun, means for defining plasma boundary form and position, means for defining the beam trajectories in the guiding channel, as well as means for interpolation the beam trajectories in the space of free moving of electrons at the technological chamber. The results of simulation for the distribution of electric field in discharge region as well as for the distribution of current density of formed electron beam on the outlet of gun are given. Also, the results for plasma boundary approximation, calculation of electrons trajectories in the guiding equipotential channel, as well as for the distribution of the density of focal beam current on the plane, located in the technological chamber outside the electron gun, are presented in the article. Main conclusion is that linear and square interpolations of electron beam boundary trajectory in the space of its free moving of electrons in technological chamber give enough accuracy for calculation the current distribution at the focal plane, and that it usually corresponded to the Gauss law of distribution. For approximation the geometry of plasma boundary for small values of discharge current the combination of exponential and linear function have been used. Obtained simulation results are very interesting and important for experts on designing and applying of electron beam equipment in industry.

xiv

Preface

N. Yeliseyeva, S. Berdnik, V. Katrich (The Radiation Characteristics of Two Coupled Vertical Dipoles with a Finite Size Screen), present a novel approach for the 3D vector problem of diffraction of the fields of two coupled vertical electric dipoles placed over an infinitely thin rectangular ideally conducting screen. On the base of the asymptotic solution to this problem by the method of the uniform geometric diffraction theory, fast algorithms and software are developed for computation of the radiation patterns and directive gains at maximum radiation, as well as radiation resistances as functions of the electric dimensions of the radiating system. It is shown that, when the removal of dipoles from screen is fixed, the appropriately chosen distance between the dipoles and the appropriately chosen dimensions of the screen provide for the symmetric patterns with high directive gain. When the dipoles currents are in quadrature, the patterns may be unidirectional. For verification of the results, the patterns and directive gains of the radiating systems are calculated using Feko software. In chapter The Analysis of Distributed Two-Layers Components in ThreeLayer Planar Structure by Yu. Rassokhina, V. Krizhanovski, V. Komarov, the authors discuss about distributed two-layer components (also known as defective ground structures), which are widely used to create filters with a wide stop-band, including matching networks for high-efficiency amplifiers. Usually, they are calculated either by equivalent circuit methods or by numerical methods. For a better understanding of physical processes in a distributed two-layer discontinuity in three-layer structures, it is desirable to develop semi-analytical methods that allow us to obtain dependences of constant propagation on parameters of structure. The transverse resonance technique (TRT) is close to such methods, which calculates the scattering characteristics of discontinuity in planar-type transmission lines in the microwave range. It is based on the solution of boundary value problems for threedimensional resonators containing, in the general case, multiplane discontinuity. Such name of this method is due to introduction of boundary conditions in the transverse with respect to the discontinuity of the direction in terms of the coefficients of reflection from the ideal boundaries (electric or magnetic). The key to effectively solving a boundary-value problem (for example, by the Galerkin method) is its algebraization method that is, the choosing a basis on which the unknown field or current components are decomposed. The chapter Planar Bandpass Filters with Mixed Couplings by A. Zakharov, M. Ilchenko, S. Rozenko, L. Pinchuk is connected with some aspects related to the designing stripline and microstrip bandpass filters with mixed coupling, including the magnetic and electric components of the interaction. It is shown that the transmission zero corresponding to mixed coupling coefficients can be shifted long the frequency axis by changing the shape of the stepped-impedance resonators. It is confirmed that N-resonator planar filters can have (N–1) transmission zeros. Designs of microstrip filters with combined coupling, which include mixed coupling and the traditionally used magnetic and electric coupling, are proposed. It is shown that the number of transmission zeros of such filters is smaller than for filters with only mixed coupling, but their designing and tuning are less labor-consuming. The data of the experiment and computer simulation are presented.

Preface

xv

D. Mayboroda, S. Pogarsky (Microstrip Monopole Antenna with Complicated Topology) present a novel approach for one of the possible constructions of planetype antenna based on microstrip resonator with radiator of complex topology, excited by coaxial line segment. The results of the study of frequency and energy characteristics of multi-band antenna based on microstrip monopole with the sectorial type inhomogeneity are presented. The main characteristics have been obtained through numerical modeling based on semi-open resonator model. The finite element method (FEM) has been chosen for numerical modeling of frequency and energy parameters. The Ansoft HFSS software was used for numerical simulations. The main antenna characteristics have been analyzed. A. Shmat’ko, V. Mizernik, E. Odarenko (Scattering of Electromagnetic Wave by Bragg Reflector with Gyrotropic Layers) solved the problem of scattering of plane wave on a ferrite 1D magnetophotonic crystal controlled by a DC transverse magnetic field. Fundamental solutions of the Hill equation with mixed boundary conditions based on the Floquet-Bloch theory are obtained in an analytical form. The dispersion equation and its roots are found explicitly. The analysis of the dispersion properties of the structures is carried out depending on the material parameters of the ferrite layers. The transmission and reflection coefficients are determined for the gyrotropic crystal with finite number of periods. Two characteristic cases are considered: positive and negative values of the effective permeability of gyrotropic layer. The expressions for spatial distribution of electromagnetic field components are determined at crystal period. The results provide a deeper understanding of the electromagnetic waves propagation behavior in multilayer media with controlled gyrotropic elements. In addition, the obtained analytical expressions simplify the analysis of wave processes in such complex media. In the chapter Theory of color constancy of multimedia images by V. Pyliavskyi, P. Vorobienko, the authors pay attention to the fact that due to the increase in the use of multimedia images (for example, in medicine, military affairs), the quality requirements of multimedia images transmitted through communication channels have increased significantly. In most cases, existing methods of transmission and correction of multimedia images do not allow obtaining an adequate original image. This is due to the complexity of solving the problem, which is a large variety of spectral characteristics of the light sources, the discrepancy of the spectral characteristics of the cameras and reproduction devices with standard / standardized requirements. A method of spectral adaptation to light sources is proposed which has several variants of adaptation with the use of an additional device for determining the spectrum of sources of light and without, and the determination of color coordinates is performed using the color temperature curve. The method of determining the coordinates of the color of the light sources is replaced by an existing graph-analytical method. A method for determining the range of colors to be transmitted has been added, taking into account system characteristics such as spectral characteristics of transmitting and reproducing devices. The theorem is proved that if the light source, the transmitting device (camera, etc.) and the reproducing device meet the set requirements, then there will

xvi

Preface

always be a method of transmitting an optical signal whose spectral composition will differ from the real one at an infinitesimal amount. We would like to sincerely thank the authors of this collection, because without their hard work of preparing good chapters, this volume would not have been successfully prepared. And last but not least, the book’s editors would like to thank everyone from the Springer Nature for their dedication and assistance in completing and finishing this large publication project on time, for supporting the highest publication standards, especially the Series editor, Prof. Janusz Kacprzyk, Polish Academy of Sciences, Dr. Thomas Ditzinger, Executive Editor, Interdisciplinary and Applied Sciences & Engineering, Ms. Varsha Prabakaran, Project Coordinator, Books Production. June 2020

Janusz Kacprzyk

Contents

Modern Challenges in Information Technologies The Main Directions of Improving Information and Communication Technologies in the Global Trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mykhailo Ilchenko, Leonid Uryvsky, and Sergey Osypchuk

3

From Big Data to Smart Data: The Most Effective Approaches for Data Analytics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andriy Luntovskyy, Larysa Globa, and Bohdan Shubyn

23

Cloud-Based Architecture Development to Share Vehicle and Traffic Information for Industry 4.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Serhat Bulut Ibrahim and Haci Ilhan

41

Complex Approach in Cryptanalysis of Internet of Things (IoT) Using Blockchain Technology and Lattice-Based Cryptosystem . . . . . . . . . . . . Lela Mirtskhulava, Larysa Globa, Nana Gulua, and Nugzar Meshveliani

55

Biometric Cryptosystems: Overview, State-of-the-Art and Perspective Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maria Lutsenko, Alexandr Kuznetsov, Anastasiia Kiian, Oleksii Smirnov, and Tetiana Kuznetsova

66

Biometric Authentication Using Convolutional Neural Networks . . . . . . Alexandr Kuznetsov, Inna Oleshko, Kyrylo Chernov, Mykhaylo Bagmut, and Tetiana Smirnova

85

Transdisciplinary Fundamentals of Information-Analytical Activity . . . Stanislav Dovgyi and Oleksandr Stryzhak

99

Ontological Fundamentals of Scientific and Education Portals . . . . . . . . 127 Stanislav Dovgyi, Oksen Lisovyi, Nadiay Gayevska, and Iurii Mosenkis

xvii

xviii

Contents

Modern Challenges in Telecommunication Technologies Data Transmission Performance Enhancement in Multi-gigabit Wide Area Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 Kirill Karpov, Veronika Kirova, Nikolai Mareev, Dmytro Syzov, and Eduard Siemens System of Solutions the Maximum Number of Disjoint Paths Computation Under Quality of Service and Security Parameters . . . . . . 191 Oleksandr Lemeshko, Oleksandra Yeremenko, Maryna Yevdokymenko, and Batoul Sleiman Conditionally Infinite Telecommunication Resource for Subscribers . . . 206 Larysa Globa, Mariia Skulysh, and Eduard Siemens Different Extrapolation Methods in Problems of Forecasting . . . . . . . . . 217 Irina Strelkovskaya, Irina Solovskaya, and Anastasiya Makoganiuk Methods for Calculating the Performance Indicators of IP Multimedia Subsystem (IMS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229 Oleksandr Romanov, Mykolaiy Nesterenko, Leonid Veres, Roman Kamarali, and Ivan Saychenko Implementation Biometric Data Security in Remote Authentication Systems via Network Steganography . . . . . . . . . . . . . . . . . . . . . . . . . . . 257 Galyna Liashenko and Andrii Astrakhantsev Wireless Systems with New Cooperative Relaying Algorithm . . . . . . . . . 274 Liana Afanasieva and Sergey Kravchuk Autonomous Unmanned Aerial Vehicles Communications on the Base of Software-Defined Radio . . . . . . . . . . . . . . . . . . . . . . . . . 289 Mykola Kaidenko and Sergey Kravchuk Estimation of the Motion Parameters of the UAVs FANET Using the Dynamic Filtering Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 303 Oleg Tsukanov and Evheny Yakornov Modern Challenges in RadioElectronics Technologies Universal Complex Model for Estimation the Beam Current Density of High Voltage Glow Discharge Electron Guns . . . . . . . . . . . . . . . . . . . 319 Igor Melnyk, Sergey Tyhai, and Alina Pochynok The Radiation Characteristics of Two Coupled Vertical Dipoles with a Finite Size Screen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342 Nadezhda Yeliseyeva, Sergey Berdnik, and Victor Katrich The Analysis of Distributed Two-Layers Components in Three-Layer Planar Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359 Yulia Rassokhina, Vladimir Krizhanovski, and Vasyl Komarov

Contents

xix

Planar Bandpass Filters with Mixed Couplings . . . . . . . . . . . . . . . . . . . 377 Alexander Zakharov, Mykhailo Ilchenko, Sergii Rozenko, and Ludmila Pinchuk Microstrip Monopole Antenna with Complicated Topology . . . . . . . . . . 394 Dmitry Mayboroda and Sergey Pogarsky Scattering of Electromagnetic Wave By Bragg Reflector with Gyrotropic Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404 Alexander Shmat’ko, Victoriya Mizernik, and E. Odarenko Theory of Color Constancy of Multimedia Images . . . . . . . . . . . . . . . . . 417 Volodymyr Pyliavskyi and Petro Vorobienko Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435

Modern Challenges in Information Technologies

The Main Directions of Improving Information and Communication Technologies in the Global Trends Mykhailo Ilchenko(B)

, Leonid Uryvsky(B)

, and Sergey Osypchuk(B)

Telecommunication Systems Department, Igor Sikorsky Kyiv Polytechnic Institute, Industrialnyi Lane 2 (Campus 30), Kyiv 03056, Ukraine [email protected], [email protected], [email protected]

Abstract. In the article an analysis of modern and promising directions of formation, processing and transmission of information are provided. Information and telecommunication technologies (ITT) in the light of global trends are considered as a single technological complex. Key trends in the development of ITT have been identified, including, in the light of the new breakthrough achievements of 2019, which are associated with the further development of ITT. Particular accent is attached to the development of mobile telecommunications (technology 5G and 6G), the development of IoT, the penetration of ITT into the development of Industry 4.0, and the integration of the global satellite Internet. The active desire of Ukraine in the structure of ITT to be at the level of world modern trends are emphasized. Keywords: Information-telecommunication technologies (ICTs) · ICTs trends · 5G · Internet · IoT · Industry 4.0 · Artificial intelligence · Satellite Internet · Cybersecurity

1 Introduction The synergy of information and communication technologies (ICTs) has an impact on all aspects of society more and more today. Back in 1982, academician V.M. Glushkov predicted a future revolutionary transformation of the existing computing means and communication technologies, and globalization of their usage. According to V.M. Glushkov [1], “…The merging of telecommunications with machine informatics (implemented in computer networks and computer centers with remote terminals) has already led to the emergence of a new term “telematics”. The most zealous apologists for telematics predict that the day is not far off when ordinary books, newspapers and magazines will disappear. Instead, each person will carry an “electronic” notepad, which is a combination of a flat display with a miniature radio transceiver. By typing the desired code on the keyboard of this “notepad”, you can (from anywhere on our planet) get any texts, images (including dynamic ones) from © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 3–22, 2021. https://doi.org/10.1007/978-3-030-58359-0_1

4

M. Ilchenko et al.

giant computer databases connected to the network that will replace not only modern books, magazines and newspapers, but also modern TVs. The progress of electronic technology, machine informatics and telematics is taking place at such a rapid pace that fiction in this area is becoming a reality literally in front of our eyes” [1, p. 537]. One of the first documents devoted to this topic in the 21st century is Okinawa Charter on Global Information Society of 22 July 2000 [2], signed by the heads of the G8 leading countries. It states: “Information and communication technologies (IT) – one of the most powerful forces in shaping the twenty-first century. Its revolutionary impact affects the way people live, learn and work, and interaction between government and civil society… There are plenty of opportunities for everyone to get an advantage from and to share between all of them”. The key component of the mentioned strategy was a continuous movement towards universal access to information and communication resources. The Internet became as a top of “paperless informatics” as the fastest growing branch of modern communications. This provision is supported by the UN in the global paper entitled “Millennium Declaration of the United Nations” adopted resolution 55/2 of the General Assembly of 8 September 2000 [3]. The document says: “We need to take steps to ensure that all people can use the benefits of new technologies, especially information and communication technologies …”. The well-known company Deloitte Global makes predictions on the latest advances in modern information technologies [3]: 1. The number of smart speakers sold is expected to exceed 250 million units by the end of the year. Smart speaker sales number will be $7 billion in 2019. This will turn them into the category of network devices with the highest sales growth rates. 2. 5G networks will appear on the market in 2019 (what had happened already); by 2020, approximately 50 operators will begin offering 5G services. 3. Approximately 40% of all television viewing by men aged 25–34 years will be provided with the opportunity to make the sports bets in the United States. 4. China will become the world leader in telecommunications networks in 2019 and is likely to maintain its leadership position in the medium term. Revenue from the sale of semiconductor devices manufactured in China will grow by 25% and amount to $120 billion. This will turn China into a player of global importance in the production and development of semiconductor technologies. 5. A new record for 3D printers. The sales of 3D printers, materials and services of large companies will exceed $2.7 billion in 2019 and reach $3 billion in 2020. This indicates a 12.5% increase in sales annually. 6. The popularity of eSports will continue to grow. The North American eSports market will grow 35% due to wide advertising, broadcast licenses and franchise sales. 7. Broadcasting remains relevant. According to Deloitte forecasts, more than 85% of the adult population in developed countries listen to the radio at least weekly (the same level as in 2018). However, this indicator will vary in developing countries. In total, nearly 3 billion people worldwide will listen to the radio weekly. 90% of the US population aged 18 to 34 will do this in 2019 [4]. Therefore, we will highlight the key areas of ICTs development in the World and Ukraine. The trends we found are based on the information and news from World Wide

The Main Directions of Improving Information and Communication Technologies

5

Web. This news is analyzed by the authors during the first half of 2019 mainly. As in the time of V.M. Glushkov, the words combination “information and telecommunication technologies” means the possibilities totality for creation and filling of information resources and information transmission by telecommunication means on local and global scales, which corresponds to the content of the modern term “infocommunications”. The following sections highlight are the key areas for ICT development that are expanded in current paper: 1. ICT: the dynamics of development in Ukraine and the world. 2. 5G Development: in 2019, 5G networks were launched for the first time. 3. The Internet of Things: key development trend and developments related to the IoT direction and the impact of ICT in this direction. 4. Industry 4.0: consider the main features of Industry 4.0 and key trends in ICT development in the world and in Ukraine. 5. Global Satellite Internet: becoming more relevant today in today’s world. 6. Artificial intelligence (AI): a part of ICT that already has a great influence in the modern world, is developing and will take an integral part in the processes of life of modern society. 7. Cybersecurity: the issues of information security and its protection during transmission and storage are also among the most pressing today. 8. New Breakthroughs: For the first time in the world in 2019, let’s focus on the scientific and practical developments that took place in 2019.

2 Global Trends in ICTs Development in the World and Ukraine 2.1 Global Digitalization 52% of World population today has no access to the Internet. Over the past 7 years the number of people connected to the global network increased by 1.5 billion people. These numbers are given in the UN report “The State of Broadband 2017: Broadband Catalyzing Sustainable Development” that was prepared jointly with the International Telecommunication Union (ITU) and UNESCO, wrote sostav.ua [5]. Most Internet users live in China. In China almost 700 million people have access to the network. Next is India with 355 million people who have access to Internet. 62% of people who do not use the Internet live in Asia and on islands in the Pacific Ocean. In Africa, 18% of people do not have access to the worldwide network. The fastest Internet in South Korea is 28.6 Mbps. The average speed in the world is 7.2 Mbps, the report analysts said [5]. Digitization of information flows and increasing the information transfer rate are the main ICTs trends. What problems the Digitalization solves? First, it is convenience – all transactions and reports are available in a couple clicks, and receipts and statements for each transaction are not lost because the mail is stored in the personal account on the site or app. Second, data rate. It doesn’t require an explanation. Just remember the case of need to come and interact with cashier, even when there is no queue of customers.

6

M. Ilchenko et al.

Thirdly, accessibility in mode 24/7/365, from anywhere in the world [6]. The global terrestrial and space resources are used for it. According to the study, current trends are associated with the rapid development of fiber optic networks and the acceleration of broadband Internet penetration will become the driving force for the broadband networks’ growth by 2025. The authors of the forecast argue that some types of fiber-optic technologies – in particular, such as Fiber to the Home (FTTH), Fiber To The Premises (FTTP) and Fiber To The Business (FTTB) will be used by 2025 to connect 59% of all broadband internet subscribers to the network [7]. Fixed and mobile communication technologies convergence and billion subscribers will use the Internet via combination of these two technologies. The new terrestrial technology platform “10X Faster Internet Speeds” will help to “accelerate” the 1 Gbps networks to 10 Gbps networks [8]. The OneWeb company launched the first six satellites for access to Internet [9]. The SpaceX company also has plans to create Satellites Network for access to Internet. Federal Communications Commission of the United States approved sending to the Space 7000 satellites. The all satellite number is going to count 12000 satellites. Ilon Musk promises finally to provide 1 Gbps average data rate on the Earth for access to Internet [9]. However, the practical application of products development “at home” in Ukraine often occurs too slowly. Consumer surveys of digital banking by KPMG show that Ukrainians are increasingly positive about digital, although personal communication is still dominant. For example, to obtain customer support, 75% of people will choose a personal communication channel and only 18% will choose an electronic one. A similar trend is observed when there is a need to buy a financial product or service – most people trust a personal communication channel [5]. At the same time, the number of IT specialists in Ukraine is growing, the quality and creativity of startups is striking, and products release to international arena is accelerated. At the end of 2018, the IT industry in Ukraine took second place in terms of exports, selling overseas products worth $4.5 billion. These figures were voiced by CEO of UNIT.City Innovation Park Maxim Yakover, reports Economic Truth. For comparison, in 2017 the volume of exports of the IT industry amounted to $3.6 billion, or 3.4% of the country’s GDP. Also, at the end of 2018, 20% of companies – world leaders in the field of software development for mobile platforms – had offices in Ukraine. There are also four thousand IT companies and more than 110 R&D centers of worldfamous international companies in Ukraine. It was previously reported that computer services in 2018 accounted for more than 20% of all exported services [10]. 2.2 5G Technology Implementation and Its Intensification 5G technology – is the name that is used in scientific studies and projects to refer to these telecommunication standards for mobile networking standards after 4G. 5G also seeks to reduce delays and reduce the electricity consumption (which is important for devices such as “Internet of Things”) compared to 4G [9].

The Main Directions of Improving Information and Communication Technologies

7

5G should provide the following features: • peak download data rates for one base station up to 20 Gbps; • download data rates up to 100 Mbps and upload data rate up to 50 Mbps for one subscriber; • possibility for subscriber units to move at speeds up to 500 kmph between base stations (e.g., high-speed train); • possibility of devices to switch between the modes of energy saving and working mode up to 10 ms; • latency up to 4 ms under favorable conditions, and up to 1 ms for dedicated connections; • improved efficiency of radio spectrum; • data transfer rate of 1 Gbps simultaneously to many users on one floor; • the possibility of up to 1 million devices per 1 km2 [11]. As of early 2018, this technology was not fully defined in international standards, but was only under development. The first 5G networks are introduced in 2019. South Korea has emerged victorious in the 5G championship race involving the United States, China and European countries. On April 3, it became possible to use the new generation network throughout the country. “South Korea will be the locomotive for promoting this network on the planet,” said Information Technology Minister Yu Yong Min. Initially, South Korea’s 5G launch was scheduled for April 5, but after rumors emerged that Verizon was launching networks in the US on April 4, three South Korean telecommunications companies SK Telecom, KT and LG Uplus had decided to postpone the release two days earlier. Therefore, South Korea established the world’s first national network 5G [12]. China launched a trial 5G in seven cities [11]. The fifth-generation network is launched in Beijing, Shanghai, Guangzhou, Shenzhen, Nanjing, Hangzhou, as well as in the new Xunan economic region. Chairman of the telecom company Wang Xiaochu said that this year the operator plans to invest about 6–8 billion yuan (895 million-1.2 billion dollars) in 5G development [13]. So, China provided the largest 5G-covering in the world [14]. In China, the city of Shanghai, 5G network covered a large area of Hongkou. Now you can access the fifth-generation network from anywhere in the area, with an area of more than 23 km. At the ceremony dedicated to providing access to 5G, the deputy mayor of Shanghai, Wu Qing, made a video call through 5G from the folding Mate X smartphone. In general, 10 thousand 5G base stations are planned to be installed in Shanghai this year. In two years, their number should reach 30 thousand [14]. Fifth-generation mobile communications have also appeared in two US cities – Chicago and Minneapolis. These networks were deployed on Ericsson equipment. But for now, only the owners of Samsung Galaxy S10, the world’s first smartphone to transmit data over networks of the new standard, will be able to use the new service [12].

8

M. Ilchenko et al.

Switzerland was the first country in Europe to launch 5G mobile communications. Fifth generation wireless technologies soon came to the UK. Italy became the third country in Europe, where the 5G mobile communications in 28 municipalities of five large cities - Milan, Turin, Bologna, Rome and Naples were launched [15]. South Africa telecommunications company plans to launch 5G commercial network [16]. 5G mobile networks promise to revolutionize their field. According to the operators, the new fifth generation will be faster than 4G tenfold [12]. The Huawei company together with the Turkish cell operator Turkcell will start project for 5G implementation based on cloud technologies [17]. The goal of collaboration is to create a global cloud network, which will make the transition to the fifth generation standard worldwide. “It will be based on commercial cloud software architecture, separation technology “user plane” and “control plane” (CUPS) and technology A/B testing (split testing). This step will be a new stage in transforming the network architecture of operators for the widespread adoption of 5G,” as stated in the message [17]. At the same time, Huawei reached 1 Gbps data rate during testing 5G network [18]. As part of the MWC 2019 Mobile Operators Convention, Huawei conducted a test of a 5G network built using 5G LampSite Pro and a 5G CPE Pro modem. For Huawei, this was the first demonstration of the 5G home network built using one of its routers. At peak times, the downlink data rate exceeded the 1 Gbps mark when using the 100 MHz bandwidth and 4T4R technology. Huawei introduces 5G LampSite technology as a powerful solution for 5G Internet access for airports, train stations, stadiums, shopping centers, subway stations and student campuses [18]. The AT&T operator performs testing in US cities the 5G networks with 1 Gbps data rates. Yet only the approved and limited users number have access to the network to use 5G services [19]. According to the data, the network from AT&T has been working locally since last year, but at one time the Internet speed was rather low – it did not exceed 195 Mbps. Now, data transmission in the operator’s network reaches gigabit speeds. “This was made possible thanks to the aggregation of four carriers with a width of 100 MHz (previously, one carrier was used, which, according to Qualcomm, could theoretically provide 625 Mbps),” the resource said. But experts determined that the relatively low data rate 5G which is achieved, is associated primarily with the restriction of frequency resources. In big cities, many operators today can use a maximum of 375 MHz [19]. At the same time, experts believe that the introduction of 5G technology contains problematic aspects also. Most operators are concerned about the implementation of the new 5G standard. Thus, more than 90% of respondents believe that 5G will increase energy costs and are interested in technologies and services that increase the efficiency of its use. This is consistent with Vertiv’s internal analysis, which shows that switching to 5G is likely to increase the overall power consumption of networks by 150–170% by 2026, with the largest increase in macro, host, and network data center areas [20].

The Main Directions of Improving Information and Communication Technologies

9

As the wireless technology industry strives to install thousands of new cellular base stations to create ultra-fast 5G networks, questions about the impact of the new technology on human health may delay their appearance. It is reported by The Fortune. The Food and Drug Administration and the Federal Communications Commission say there’s nothing to worry about. Most studies do not associate radio signals from mobile phones and cell towers with disease. However, some experts have found increased risks from cellular networks, and further studies will be conducted over the next few years. According to HSBC Securities analyst Sunil Rajgopal, the World Health Organization’s International Agency for Research on Cancer in 2011 classified cellular radio waves as a possible carcinogen. “The race for 5G deployment may be suspended due to health problems associated with radio frequencies. These problems were about the same as with mobile phones, but a number of regulatory/community initiatives have been taken recently that require delays or direct 5G deployment bans,” – Rajgopal commented [21]. Ukraine has no licensed frequencies today to launch 5G. Thus, the introduction of the fifth-generation network in Ukraine will be possible not earlier than in 3–4 years. Currently available standards of 3G and 4G must meet the demand of people on the Internet during the next five years [22]. Nonetheless, the Kyivstar cell operator is ready to implement 5G Communications in Ukraine [23]. On April 18, Minister of Infrastructure in Ukraine Vladimir Omelyan announced the start of testing the fifth-generation communications technology and announced that 5G licenses would be put up for auction in 2020. A month later, ex-president Petro Poroshenko signed a decree on a schedule for launching fifth-generation communications in Ukraine in 2020. Statements by Ukrainian officials regarding 5G sound in unison with their colleagues from leading countries of the world [24]. The ukrainian telecommunications market experts named the following main regulatory and economic factors that affect the successful 5G deployment in Ukraine [25]: • not transparent government strategy of 5G implementation, which considers the position of cell operators, business and market needs; • availability of comprehensive procedure for 5G frequencies licensing for 5G development and launch; • consumers and markets readiness, including the availability of affordable devices that support 5G technology; • technical readiness of mobile networks, the availability of appropriate equipment and software for deploying fifth-generation communications; • real demand for business services and solutions that require 5G, because this standard is primarily targeted to industrial and infrastructural b2b-market needs, and not to individual subscribers. The main direction of this technology – it is production. “5G – is a technology not for people, but for industrial use mainly, for new areas of the developing world, including the Internet of Things …” [22]. Thus, according to experts, 5G will be much demanded on the IoT, cloud infrastructure and the virtual reality markets [23].

10

M. Ilchenko et al.

2.3 Development of the Internet of Things (IoT) Technology IoT is the next step in the ITT technological horizon. “It is obvious that the rapid development of this technology in the near future will fundamentally change not only our everyday habits, but also business processes, the functioning of the state and society,” – the Ukrainian Association says on the Internet. Businesses also consider the Internet of Things a technology that cannot be dispensed with in the future. Sergey Fedchun, director of the IT services department of Metinvest Digital, recently said in an interview with our publication, IoT technologies are indispensable for companies building modern technological facilities: “IoT is a new approach to building automation systems. It is new primarily in relation to the collection and use of information: now the data is used not only for process control, but also for analytics and subsequent forecasting. Thanks to the Internet of Things, organizations can significantly improve efficiency while saving time and resources,” – the expert explains [26]. There were several main areas in IoT technology which got the greatest development in 2018 [27]: • • • •

Connected cars (Volvo, Ford, Volkswagen, Porsche); Global changes in funding (estimated at hundreds of dollars billions); Artificial Intelligence; Security.

The European market is predicting a steady growth of IoT market. The European IoT market revenue in 2019 increased by 19.8%, to 171 billion dollars [28]. The bulk of the projected costs will be in Western Europe, and the leader will be Germany, which will provide revenue in 2019 of $35 billion. Together, the countries of Central and Eastern Europe will bring 7% of the total European spending on IoT. The main industry consumers of IoT solutions in 2019 will be individual industries ($20 billion), utilities ($19 billion), retail ($16 billion), and transportation services ($15 billion). In the latter case, two-thirds of the costs will be for the monitoring of cargo transportation and logistics solutions. The best average annual growth will be achieved by retail (CAGR 18.5%), healthcare (17.9%), local and central authorities (17.1%). The best investment volume (over $32 billion in 2019) will be shown by the consumer segment, which is actively developing solutions for smart homes, personal finance, and connected vehicles. Hardware, including various modules and sensors, will remain the largest market segment IoT with revenue in 2019 of $66 billion. Services will have about $60 billion. Software will generate $35 billion [27]. April 9 worldwide celebrates the IoT Day. For the first time in the history of Ukraine, we can say that this holiday is not a foreign holiday to Ukraine [28]. The “Lifecell” mobile operator and the “IoT Ukraine” company built the IoT network during 9 months in three Ukrainian cities: Kiev, Lviv and Kropyvnytskyi. Ten companies from various sectors in these cities already use IoT solutions built based on this IoT network to improve the efficiency of companies’ activity [29].

The Main Directions of Improving Information and Communication Technologies

11

In Kiev, Lviv and Kropyvnytskyi, the Internet of things the LoRa network was built and began to operate. Now, the network is already used by about a dozen companies from various industries, among which are suppliers of gas, water and electricity. The Internet of things network, covering about 90% of Kiev, Lviv and Kropyvnytskyi, is a joint project of the digital operator Lifecell and IoT Ukraine. Since the end of July last year, as part of a joint project, about 80 base stations have been installed in three regions. Now, among the companies that have already connected to the network or are testing its capabilities - Kirovogradgaz OJSC, KP Lvivvodokanal, Lvivoblenergo CJSC, Art Mall, Auchan Ukraine hypermarket, Pandabox logistics company and others. Among the suppliers of devices that ensure the functioning of the IoT network, not only world-wide (Libelium, Abeeway, ORION M2M), but also Ukrainian developers – Infomir TM JOBBY and Gross. The platform that is responsible for the IoT networks operation is based on LoRaWan technologies – Actility and Cisco [30]. NB-IoT (NarrowBand IoT, or NarrowBand Internet of Things) – a licensed cellular standards-based LTE technology to exchange data between devices. NB-IoT devices are able to communicate with each other at a dedicated frequency of 1800 MHz. At the same time, devices can get stable access to the network even in hard-to-reach places, for example, in basements and elevator shafts. Deployment of NB-IoT technology is based on the LTE network, while the network capacity is much higher than in the voice network – up to a thousand devices can be connected to one base station [31]. The Lifecell operator announced the start of the deployment of a working NB-IoT network for smart devices developed by PJSC Kyivgaz, which will allow you to remotely record the readings of gas meters, monitor the stability of gas supply and timely prevent possible gas leaks. The project is supported by Ericsson, Odine Solutions and Affirmed Networks. According to Lifecell, the first stage of the deployment of the Internet of things network using NB-IoT technology in the 1800 MHz band has already been completed. This technology will allow connecting tens of thousands of “smart” devices to each base station. Kievgaz specialists developed and installed devices for remote transmission of meter readings (adapters) to gas consumption meters, which automatically read and transmit data around the clock via the NB-IoT Lifecell network to the Kievgaz server in billing and monitoring systems [32]. The mobile operator “Vodafone” successfully tested its own NB-IoT network in Kiev. NB-IoT devices can communicate with each other on dedicated 1800 MHz band. Sensor can get a stable network access even in tight spaces, such as basements and elevator shaft. Thousands of devices can connect to one base station. At the same time, the company used special modules for smart metering with automated transmission of readings, which have ultra-low power consumption, both in standby mode and in data exchange mode. By connecting such devices to the NBIoT network, it is possible to achieve the economical use of resources and increase the battery life up to 10 years [31]. The “Kyivstar” mobile operator also builds its own NB-IoT network for clients of IoT applications [33]. Thus, according to experts, IoT offers great perspectives for improving the people life quality in all industries and contributes to economic growth [32].

12

M. Ilchenko et al.

However, one should not forget about the threats at the stage of introduction of IoT technologies. Torsten Prues, Senior Unified Communications Specialist at NAIT IT, talks about data security in the Internet of Things (IoT) era. The source provides security recommendations such as changing default passwords, using guest accounts, updating software, and others [34]. 2.4 Ukraine as the “State in Smartphone” V.M. Glushkov’s foregoing prediction [1], his proposal for a never-realized project of a Country-Wide Automated System (CWAS) at the present stage yet, find their practical implementation of individual subsystems in many countries of the world, in particular, in Ukraine, due to new technological possibilities of ICTs. President of Ukraine Volodymyr Zelensky has signed a decree “On Some Measures to Improve Access of Individuals and Legal Entities to Electronic Services”, aimed at streamlining and ensuring transparency of the work of state registers, development of modern electronic identification means and introduction of priority e-government services for citizens and businesses [35]. The electronic interaction system called “Trembita” began to operate in Ukraine since May 22, 2019. It is designed for data exchange between public authorities without human intervention, which speeds up the process of providing public services, reduces mechanical errors and corruption risks. In other words, it eliminates the need to queue in one state authority to take a document, which must then be taken to another state authority. The State Agency for Electronic Governance has announced the launch of the system. The large-scale implementation of Trembita gives rise to irreversible changes in the interaction of state bodies with each other, as well as in the interaction of business and citizens with the state. After all, it is the speed and transparency of state processes, which translates into time savings and rational use of budgetary funds [36]. The 125 public services are now available at the governmental portal and are fully automated. “Public procurement, privatization, reports on technical assistance and stateowned enterprises – all these services are available online for every Ukrainian citizen”. “Our goal – country in the smartphone,” – the president V. Zelensky said [37]. According to the decision makers, “we have the certain initial steps for 2019 and concrete targets to 2024. In particular, the plan involves overcoming major digital barriers – such as electronic identification. This year we start the MobileID project, introduce SmartID (Electronic Signature mobile app), and create a convenient and affordable eID Ukrainian network services. By 2024, 75% of population should be able to use reliable and secure means of eID, and then Ukrainians will forget how face to face official government people looks like, and where those are located, because over 90% of country governmental services will be operating online” [38]. 2.5 Industry 4.0 Development Based on ICTs Impact Industry 4.0 (Industry 4.0) is a leading trend “fourth industrial revolution” taking place in front of our eyes. We live in an era end of the third, the digital revolution that began in

The Main Directions of Improving Information and Communication Technologies

13

the second half of the last century. Its characteristics – the development of information and communication technologies, automation and robotics manufacturing processes, the latest scientific technology advances implementation into the practice of creating modern and promising telecommunication facilities [39–41]. Industry characteristics 4.0 – a fully automated production, where management of all processes is carried out in real-time and adapts to changing external conditions. Cyber-physical systems create virtual copies of the physical world objects, control the physical processes and make decentralized decisions. They can be combined into a single network, interact in real time, and learn for self-tuning. Internet technologies take the important role by providing communication between staff and machines. Enterprises produce products according to individual customer requirements, optimizing production costs [42]. The fourth industrial revolution, or “Industry 4.0” will be an important topic in 2019, as IoT News wrote [43]. According to Senior Vice President of Intelligent Solutions Group, John Stone, Artificial Intelligence and Machine Learning for John Deere will be as essential as the engine and transmission. Technologies association such as Artificial Intelligence, Machine Learning, and even blockchain, put the “Industry 4.0” forward. Proponents of the Industry 4.0 point to a much more fundamental changes caused by a combination of new technology with business models and processes. Ultimately this will lead to the fact that companies are changing the way of businesses value creation, income generation and communication with its customers [44]. At the same time, only 2% of industrial equipment suppliers are using IoT technology, 47% do not anticipate implementation opportunities in IoT systems, and 51% do not know about IoT opportunities for enterprises in Ukraine today [44]. 2.6 Strengthening Cybersecurity – The Priority in ICTs Implementation Cyber security – one of the problems that await humanity in 2019 year. It’s recorded in the report of the World Economic Forum in Davos. The issue of cybersecurity will always be relevant, because the hacking tools are constantly evolving. The general vector of cyber security will move in the direction of making artificial intelligence, which enables real time to proactively provide the necessary protection [45]. In two years, the global cybersecurity market will exceed $200 billion. By 2021, global cybersecurity solutions will grow to $202.3 billion. According to the consulting company Frost & Sullivan, this is almost twice as high as in 2016 ($122.4 billion). As a result, the annual rate of market share will be about 10.6%. The largest segments of the market in the forecast period will be information and communication technologies, energy, healthcare, industry and the financial sector. The highest growth rates are expected in areas such as mobile security, secure cloud storage, potential threat analysis and cyberattack prevention [46]. Cybersecurity of connected home appliances is becoming an increasingly serious problem. People trust their personal data to a growing number of devices and services. Products and devices that have traditionally operated offline now connect to the Web

14

M. Ilchenko et al.

and need protection from cyber threats. Poorly protected products threaten consumer privacy, and some devices are used to launch large-scale DDoS attacks [47]. Without solutions implementation for this problem, the advanced ICT technologies implementation won’t be possible. F-Secure cybersecurity experts released a report claiming that history will remember 2018 as a “turning point” in the threats of the Internet of things. In 2018, the company recorded a double increase in the number of attacks on Internet of things devices compared to the previous year. The rapid spread of connected devices makes IoT a more attractive target for hackers. At the end of 2018, F-Secure revealed a huge surge in the number of threats aimed at open Telnet ports. This is the same attack used by the infamous Mirai botnet. Experts note that most of the devices of large manufacturers have good protection, but millions of (usually cheap) devices fill the market from lesser-known brands, which are often the most vulnerable. These are usually devices such as webcams and routers, which can be especially dangerous when hacked. They are compromised using simple means, for example, default passwords. Experts said the number of attacks on IoT devices has doubled over the last year. “Weak passwords, known vulnerabilities, lack of upgrades. We repeat the same mistakes as in the 90s. But now there is no excuse,” – said adviser F-Secure Tom Haffni. “Any device that can connect to the Internet – potentially at risk of cyberattack,” – the Interpol report mentions in February 2018 [48]. The FBI said that “routers, wireless radio lines, timers, streaming audio/video devices, Raspberry Pis, IP cameras, DVRs, satellite dishes, garage door openers and network-attached storage devices” could be hacked for use their computing power. This processing power allows the use of botnets, such as Mirai, which can lead to a serious violation of the services of a competitor in business, a conflict in the infrastructure of another state, or simply to a general violation. Compromised devices can also be used to mine cryptocurrencies [48]. Therefore, the Technical Committee of the European Institute for Telecommunication Standards (ETSI) on cybersecurity has already issued the cybersecurity standard on the Internet of Things TS 103 645 [47]. British authorities have issued a guide to the security of IoT devices. Manufacturers “IoT Security Guide” describes the steps required to ensure the security of home smart devices. As part of a government initiative to enhance the security of home smart devices, the Department of Digital Technology, Culture, Media and Sports and the UK National Cybersecurity Center have prepared a guide for manufacturers to protect their products from common cyber-attacks [49]. It is recommended to follow a few rules that will help protect your data from hacking: • • • • •

Use a password manager; Be careful what you share on the Internet; Do not click on links; Be neat, knowledgeable and careful; Use of cloud technologies with high security;

The Main Directions of Improving Information and Communication Technologies

15

• Remove your critical data from social networks; • Protect your online surfing, etc. [50]. In Ukraine, the state took the first step: In 2018 came into force the Law of Ukraine “Basic principles of ensuring cybersecurity in Ukraine” [51]. Government defined the main concepts with the prefix “cyber”, objects of critical infrastructure (including those that are not related to IT, such as water utilities, for example) and the vector of development of the regulatory framework for cybersecurity towards harmonization with the European Union and NATO standards. Given the specifics of our market, experts identify three main tasks that Ukraine should take care of in 2019 regarding cybersecurity: • Availability of proficient and trained personnel; • The presence of companies responsible for cybersecurity issues; • Giving appropriate attention to cybersecurity issues and assessing this issue from a risk perspective [45]. 2.7 Artificial Intelligence Applications Artificial intelligence become more sophisticated in recent years, as more and more companies began to use predictive algorithms and other automated methods in various disciplines. The number of scientific papers on the subject has increased dramatically, but officials have mentioned the technology in more than 70 sessions of Congress. Because startups with artificial intelligence beat all world records. According to a new PwC report and CB Insights, venture capital companies II last year increased by 72%, reaching a record 9.3 billion dollars [52]. Regarding network technologies, artificial intelligence helps to Facebook to detect and remove prohibited content in 96.8% cases. According to Facebook, artificial intelligence and machine learning help to significantly reduce the number of prohibited contents on the social network. Facebook is also using artificial intelligence to identify messages, personal ads, photos and videos that violate the rules on advertising and sale of prohibited goods such as drugs and firearms. Artificial intelligence helped identify 65% of more than four million similar messages deleted from Facebook every quarter, compared with 24% a little over a year ago and 59% in the 4th quarter of 2018. In the first quarter of 2019, the company said it had acted on some 900,000 drug-related publications, of which 83.3% were detected using artificial intelligence. During the same period, Facebook also discovered and deleted about 670,000 firearms publications, of which 69.9% were processed before moderators or users could encounter them [53]. Some groups of scientists from various universities in the United States pay attention to the threats that the emergence of universal artificial intelligence potentially carries. The combination of computing power of machines and human intelligence will allow a strong AI to learn, solve complex problems and improve themselves. And engage in tasks for which no one was preparing them. So, AI can turn into an artificial superintelligence (ISI). According to some experts, its appearance is possible in the period from 2029 to the end of the century, according to Science Alert [54].

16

M. Ilchenko et al.

A study by the Center for the Governance of AI, Future of Humanity Institute, University of Oxford showed that American citizens generally do not fully trust artificial intelligence and want to introduce AI regulation at the legislative level [55]. US President Donald Trump instructed federal agencies to improve the country’s capabilities in the field of artificial intelligence and help people whose work is replaced by automation. The Order creates the American AI Initiative, which will focus on five areas: • • • • •

Research and development Resources Standards Training of employees International connections

At the same time, Daniel Castro, vice president of the Data Innovation Center, reports that China has already surpassed the United States in some indicators related to AI research. For example, since 2014, he has published more scientific articles than in the United States [56].

3 Perspective Directions of Global ICTs Trends in Near Future 3.1 ITT’s Role in Launching 6G Technology Federal Communications Commission (FCC) began preparations for research and development in the “sixth generation” networks and unanimously voted for the opening of a new frequency segment for 6G services. 6G experiments and further use of sixth generation networks are planned to operate in frequency range from 95 GHz to 3 THz (Terahertz). In theory, 6G networks data rate transmission can be 10 times higher than 5G has. If the task of the fifth-generation networks is to provide users with high speed throughput and minimum delay, the 6G goal is to “unite the whole world in one touch” [57]. The Nokia CEO Rajiv Suri at a recent exhibition in “Barcelona Mobile World Congress” (MWC) 2019 suggested that “to create any products related to 6G, the time has not come yet, but 6G research is just right … Standardization is expected no earlier than in 2028” [58]. At the same time, American telecom regulator opened an experimental frequency resource for testing 6G technology. This frequency range takes the band between 95 GHz and 3 THz, while the 5G range consumes the band from 6 to 100 GHz [59]. China has set up a national 6G communications research and development working group and an expert group to promote the development of 6G technology. The meeting, with the participation of several Chinese ministries, marked the official launch of research and development in the field of 6G technology in the country [60].

The Main Directions of Improving Information and Communication Technologies

17

3.2 The Global Satellite Internet as a Promising Technology of the ICTs According to a Web Foundation study, currently 3.8 billion people on Earth are still denied Internet access. Mostly they are residents of remote villages and poor urban areas. People without Internet access have fewer opportunities for education and work, they are cut off from important social discussions and digital public services [58]. According to the OneWeb project, it’s planned by 2021 to cover the Earth by affordable communication by building the first truly global communications network with high speed broadband access. Based on the fourth quarter of 2019, OneWeb relies on more than 30 satellites at the same time to create an initial orbital group of 650 satellites to ensure full global coverage. The company will offer satellites as their needs grow [61]. OneWeb plans to send 900 satellites into space to make the most powerful groups to provide broadband access to the Internet worldwide [62]. SpaceX is also planning to create a network of Internet satellites. The US Federal Communications Commission has already authorized the launch of more than 7,000 satellites. In total there should be 12,000. Elon Musk promises in this way to provide the whole Earth with Internet at an average speed of up to gigabits per second [59]. Amazon intends to send into space more than 3,000 satellites to provide Internet access to residents of remote regions of the planet. The company has already requested International Telecommunication Union for approval to send a group of satellite to the orbit, – the initiative is called “Project Kuiper”. Amazon plans to bring into low Earth orbit 3236 satellites, 784 of which will be placed at an altitude of 367 miles (590 km), 1296 satellites – at an altitude of 379 miles (610 km), and 1156 satellites – the 391-mile (630-km) orbit [63]. Elon Musk showed the dense configuration of 60 satellites for the Internet. SpaceX company launched into Earth’s orbit 60 telecommunication satellites Starlink in late May 2019. All 60 Starlink running satellites operating in normal mode [64]. The Lockheed Martin company has developed a new system “LTE via satellite” designed to provide communication services to clients who are in remote regions, in particular – in areas where there is no coverage of cellular networks for communication with the boats at sea, and provide communication during natural disasters – such as hurricanes, fires, earthquakes, floods and volcanic eruptions [65]. The US NASA space agency will conduct the first test of new communication system that will work in X-rays [66]. Ukrainian scientists have teamed up to accelerate work on a new telecommunications satellite. Leading scientists of various technical universities of Ukraine have begun work on the design and construction of the next generation spacecraft, which will be able to provide satellite broadcasting of all Ukrainian TV channels, including in ultra-high definition format. The working title of the project is “Pribulets-1”. For its creation, the development group has already received about 777 billion New Zealand dollars, which will be mastered as soon as possible. Scientists promise to complete the project within the next 2–3 years, then it will remain to send the satellite into orbit, but this part of the work will no longer be their responsibility [67].

18

M. Ilchenko et al.

3.3 Moving to New TV Broadcast Formats The Project “New TV” YouTube prepares for world domination. Internet resource YouTube excels not only cable networks and is the best source of entertainment on TVs for millennials (people born in the late 80’s–early 90’s). In less than a year, the service’s video viewing on “stationary” TVs increased by up to one third – up to 250 million hours of viewing per day. This was reported by Variety, with reference to a statement by YouTube CEO Susan Wojcicki made at the Brandcast marketing event in New York. The plans of the service are to deal with the challenges of the 21st century and attract even more young audiences and advertisers. On all YouTube platforms, in general, the number of consumers between 18 and 49 in the US per week is on average higher than all cable television networks combined, as stated by Allan Thygesen, president of Google in America. In fact, Taigesen cited a study that Google ordered from Nielsen a year ago for statistics. YouTube surpasses not only cable, but also broadcast networks as the preferred source of entertainment on TVs among millennials [68]. The Ukrainian Media Group “Our Media” launched the Media Channel “OUR HDR”, with 4K resolution. Broadcasters testing this channel via the Ukrkosmos satellite. It is Ukraine’s first media company that decided to broadcast their channels in three versions – a regular digital (SD), HD and 4K [69]. 3.4 Threats and Social Challenges of the Internet The Internet creator Tim Berners-Lee calls to fight the threats facing the Global Network. He outlined three challenges risen before him. Two years later they are still relevant [70]. The first danger is that “we lost control over our personal data”. He noted that many websites operate on a business model that allows users to have free content in exchange for their personal data, which allows the collection of a wealth of personal information, which in turn can make the Internet a dangerous place to store anything – any privacy, or to discuss important but sensitive topics. The second problem is that the Internet is “too easy to spread misinformation”. Noting that “most people find news and information on the Internet only through a handful of sources,” he noted that websites can easily and quickly spread false stories and false information. Thirdly, political advertising “needs transparency and understanding” on the Internet. He noted that such advertising could easily be used to manipulate people who see it.

4 New Breakthroughs of 2019 that Will Impact ITs Development 4.1 Spin Lasers Spin lasers can speed up fiber optic communication 5 times or more [71]. Engineers at the University of Ruhr (RuhrUniversität Bochum) developed a fundamentally new concept for fast data transmission over fiber optic cable. It is not based on the modulation of light intensity, as is done in modern systems, but it is based on its polarization. The tested lasers generate light with circular polarization obtained by imposing two perpendicular (linearly) polarized waves. The frequency of these oscillations could spin and adjust it higher than 200 GHz.

The Main Directions of Improving Information and Communication Technologies

19

4.2 Quantum Internet Physicists developed a device for sending and receiving bits of information encoded in photons of infrared radiation (qubits). In 2020, there are plans to launch a demonstration network that includes four cities in the Netherlands. This network may be the first of its kind and will transmit qubits between any two network nodes [72]. 4.3 Providing Mobile Underground and Under Water The antenna is designed to provide mobile communications under the ground and under water [73]. Much longer wavelengths of VLF-range allows to penetrate deep into the earth to many meters and overcome distances of miles thousands over the horizon. This solution is the key to achieve data rate of more than 100 bits per second, what is enough to send plain text. 4.4 Energy-Independent Provision for ICTs There was a tendency to create energy-independent means for information transmission in 2019. For the first time, a battery-free mobile phone has been introduced to the world [74]. Device takes energy from the air. In addition, phone converts the signal received from the base station into energy. Another example is a microbial bio-battery that can provide energy to IoT devices [75]. Scientists at Bingampton University in New York developed a suitable power source for IoDT devices – a microbial cell battery. The Internet of Disposable Things (IoDT) is a new paradigm whereby small and cheap wireless sensors connect to virtually any device type to provide up-to-date information over the Internet [76]. For example, sensor can be attached to a food package to control the freshness of food inside. These devices are designed to work for a very, very short period and then are disposed.

5 Conclusion 1. Nearly 40 years passed since the brilliant foresight of V.M. Glushkov [1] regarding the powerful synergy of integration the information and telecommunication technologies. At the same time, global experience confirmed the importance of ICTs for the development of all human activity spheres. Today, there is a high positive trend of ICTs development, in the last year regarding global digitalization, introduction of 5G technology, IoT use, ICTs penetration into the Industry 4.0 development process. 2. The following global ICTs trends should be considered as promising in near future: start creating 6G technology; practical steps to implement the global satellite Internet; development of new TV broadcast formats; considering threats and social challenges of Internet. New breakthrough technologies related to ICTs development are of interest to professionals of all levels, such as spin lasers, quantum internet, mobile communications underground and underwater, microbial bio-batteries for disposable Internet, and artificial intelligence.

20

M. Ilchenko et al.

3. Positive changes of the state’s attention to digitization of all Ukraine activity spheres are illustrated by the Presidential Decree provisions “On Some Actions to Improve Access of Individuals and Legal Entities to Electronic Services”, which aims at streamlining and ensuring transparency of the state registers work, modern electronic identification means development and e-government services introduction for citizens and business [35]. The tasks implementation of this Decree is based on the information and telecommunication technologies application within the framework of new Ukraine approach for IT and telecom industries formulation and functioning [77].

References 1. Glushkov Fundamentals Bezbumazhnoy informatics, Ch. Ed, fyz.-mate. Science, Moscow (1982). 552 p. 2. Okinawa Charter on Global Information Society/Per. Official, 22 July 2000. Source: Japan Foreign Ministry official site. https://www.mofa.go.jp/policy/economy/summit/2000/docume nts/charter.html 3. Millennium Declaration Organization Obedynennyh nation. Accepted resolutions 55/2 Heneralnoy Assembly from 8 September 2000. https://www.un.org/ru/documents/decl_conv/ declarations/summitdecl.shtm 4. https://tech.liga.net/technology/novosti/kolonki-5g-3d-printery-i-kitay-prognoz-budesc hego-ot-deloitte/ 5. https://gazeta.ua/articles/science/_povidomili-de-nazemli-ne-koristuyutsya-internetom/ 793565 6. https://uazmi.org/news/post/hg2o6K4dakGhf0CqrQya1w/ 7. https://mediasat.info/2019/03/12/global-fixed-broadbandsubs/ 8. https://mediasat.info/2019/01/10/new-technologyplatform-10-gigabit-internet-speeds/ 9. https://my.ua/news/cluster/2019-02-28-pervye-sputnikisistemy-internet-dostupa-onewebvyvedeny-naorbitu/fddab0a8-7709-2779-de8c-98edba376614 10. https://korrespondent.net/business/companies/4060086-IT-yndustryia-v-ukrayne-vyshla-vlydery-po-eksportu/ 11. https://uk.wikipedia.org/wiki/5G/ 12. https://ua.korrespondent.net/business/web/4084076peremozhets-honky-pivdenna-koreiapershoui-vvela-5G/ 13. https://mediasat.info/2019/04/23/kitaj-5g/ 14. https://mediasat.info/2019/04/02/5g-kitaj-2/ 15. https://mediasat.info/2019/02/21/huawei-i-turkcell/ 16. https://mediasat.info/2019/02/27/huawei-5g/ 17. https://internetua.com/5g-ot-at-t-dostig-gigabitnoi-skorosti/ 18. https://mediasat.info/2019/06/07/vodafone-italia-5g/ 19. https://mediasat.info/2019/02/28/yuar5g/ 20. https://ko.com.ua/perehod_na_5g_znachitelno_uvelichit_jenergopotreblenie_mobilnyh_ setej_128028 21. https://news.ua/technologies/razvertyvanie-5g-mogut-priostanovit-iz-za-riskov-dlya-zdorov ya-cheloveka/ 22. https://lantorg.com/article/ukraine-rano-zhdat-5g/ 23. https://mediasat.info/2019/04/04/kizvstar-5g/

The Main Directions of Improving Information and Communication Technologies

21

24. https://www.epravda.com.ua/rus/publications/2019/06/6/648497/-19+1 25. https://zn.ua/ECONOMICS/eksperty-nazvaliosnovnye-faktory-dlya-razvertyvaniya-5g-vukraine319520_.html 26. https://internetua.com/den-interneta-vesxei-s-csemukraina-vhodit-v-novuua-tehnologicse skuua-realnost/ 27. https://internetua.com/eksperty-internet-vesxei-2019sleduuasxii-shag-za-gorizont/ 28. https://ko.com.ua/evropejskij_rynok_interneta_veshhej_zhdet_uverennyj_rost_127650 29. https://itc.ua/news/lifecell-v-partnerstve-s-iot-ukrainerazvernul-set-interneta-veshhej-vtreh-oblastnyh-tsentrahukrainy/ 30. https://blog.ukrnames.com/novosti/v-treh-ukrainskihgorodah-nachala-deystvovat-set-interz eta-veshhey/ 31. https://itc.ua/news/vodafone-uspeshno-protestiroval-vkieve-sobstvennuyu-set-interneta-ves hhej-nb-iot/ 32. https://ko.com.ua/lifecell_razvertyvaet_set_nbiot_dlya_umnyh_schetchikov_kievgaza_1 27347 33. https://ko.com.ua/kievstar_stroit_dlya_klientov_seti_interneta_veshhej_na_baze_nb-iot_ 127751 34. https://internetua.com/eksperty-kak-ne-stat-jertvoi-interneta-vesxei/ 35. https://www.ukrinform.ua/rubric-society/2750471-prezident-pidpisav-ukaz-pro-rozvitokeposlug.html 36. https://itc.ua/news/v-ukraine-zapushhena-sistemaavtomaticheskogo-obmena-dannymi-mez hdugosudarstvennymi-organami-pod-nazvaniem-trembita/ 37. https://ru.espreso.tv/news/2019/05/31/gosudarstvo_v_smartfone_uzhe_rabotaet_kubyv _otvetyl_na_upreky_zelensko go 38. https://tech.liga.net/technology/novosti/u-zelenskogorasskazali-kak-budut-perevodit-gosusl ugi-v-onlayn 39. Zakharov, A., Rozenko, S., Ilchenko, M.: Two types of trisection bandpass filters with mixed cross-coupling. IEEE Microw. Wirel. Compon. Lett. 28(7), 585–587 (2018) 40. Zakharov, A., Litvintsev, S., Ilchenko, M.: Trisection bandpass filters with all mixed couplings. IEEE Microw. Wirel. Compon. Lett. 29(9), 592–594 (2019) 41. Zakharov, A., Rozenko, S., Ilchenko, M.: Varactor-tuned microstrip bandpass filter with loop hairpin and combline resonators. IEEE Trans. Circuits Syst. II Exp. Briefs 66(6), 953–957 (2019) 42. https://www.it.ua/knowledge-base/technologyinnovation/industry-4/ 43. https://internetua.com/kak-budet-razvivatsyaindustriya-4-0-v-2019-godu 44. https://tech.liga.net/technology/novosti/v-ukrainetolko-2-postavschikov-ispolzuyut-IoT 45. https://tech.liga.net/technology/article/naperegonki-shakerami-chego-jdat-po-kiberbezopas nosti-v-2019-m 46. https://ko.com.ua/mirovoj_rynok_kiberbezopasnosti_cherez_dva_goda_prevysit_200_ mlrd_doll_127874/ 47. https://iot.ru/bezopasnost/etsi-vypustil-pervyyglobalnyy-standart-bezopasnosti-dlya-potreb itelskogo-iot/ 48. https://internetua.com/eksperty-kolicsestvo-atak-naustroistva-interneta-vesxei-vyroslo-vd voe/ 49. https://www.securitylab.ru/news/495955.php 50. https://ubr.ua/ukraine-and-world/technology/kiberhihiena-kak-zashchitit-svoi-dannye-v-int ernete-3880491/ 51. https://jurliga.ligazakon.net/news/170010_zakon-okiberbezopasnosti-vstupil-v-silu/ 52. https://internetua.com/startapy-s-iskusstvennymintellektom-pobili-vse-mirovye-rekordy/ 53. https://3dnews.ru/988061/

22

M. Ilchenko et al.

54. https://internetua.com/eksperty-iskusstvennyi-intellekt-mojet-unicstojit-cselovecsestvo/ 55. https://www.liga.net/world/opinion/nastoyaschie-zakony-robotehniki-pravo-i-iskusstve nnyy-intellekt/ 56. https://tech.liga.net/technology/novosti/tramp-poruchil-usilit-rabotu-ssha-nad-iskusstve nnym-intellektom/ 57. https://internetua.com/v-ssha-uje-gotovyatsya-kepohe-6g/ 58. https://portaltele.com.ua/news/events/6g-mechtat-nezapretish.html 59. https://mediasat.info/2019/03/20/us-6g/ 60. https://gordonua.com/news/worldnews/v-kitae-nachali-izuchenie-i-razrabotku-tehnologii6g-1412596.html 61. https://tech.liga.net/telecom/novosti/oneweb-planiruetk-2021-godu-pokryt-zemlyu-dostup noy-svyazyu/ 62. https://mediasat.info/2019/03/18/oneweb-investicii/ 63. https://mediasat.info/2019/04/05/amazon-sputniki/ 64. https://tech.liga.net/technology/novosti/ilon-maskpokazal-plotnuyu-kompanovku-60-sputni kov-dlya-interneta/ 65. https://itc.ua/blogs/vse-60-zapushhennyh-sputnikovstarlink-rabotayut-v-shtatnom-rezhime/ 66. https://www.rbc.ua/rus/news/nasa-oprobuet-novuyu-sistemu-svyazi-1554014945.html 67. https://mediasat.info/2019/04/01/pribulec/ 68. https://twnews.com.ua/ua-news/six-p-novyi-televizor-youtube-gotovitsia-k-mirovomu-gos podstvu/ 69. https://tech.liga.net/technology/article/oligarhi-ne-pridelah-kak-v-ukraine-poyavilos-televi denie-v-4k 70. https://twnews.com.ua/ua-news/tvorets-internetuzaklikaie-borotisia-iz-zagrozami-shcho-sto iat-pered-vsesvitnoiu-merezheiu/ 71. https://ko.com.ua/spinovye_lazery_mogut_uskorit_optovolokonnye_kommunikacii_v_5_ i_bolee_raz_128358 72. https://tech.liga.net/telecom/article/kubit-vmesto-bitakak-budet-rabotat-kvantovyy-internet/ 73. https://mediasat.info/2019/05/03/super-antenna/ 74. https://hi-tech.mail.ru/news/telefonkotoromu bez batarejki/imageset/12359759 75. https://strichka.com/article/92202572/ 76. https://read.nxtbook.com/ieee/spectrum_int/spectrum_int_december_2018/the_internet_of_ disposable_th.html 77. https://tech.liga.net/technology/interview/didjitalstrateg-zelenskogo-za-kajdym-reestromest-smotryaschiy-otkriminal

From Big Data to Smart Data: The Most Effective Approaches for Data Analytics Andriy Luntovskyy1(B)

, Larysa Globa2(B)

, and Bohdan Shubyn3(B)

1 BA Dresden University of Cooperative Education (Saxon Study Academy),

Hans-Grundig-Street 25, 01307 Dresden, Germany [email protected] 2 National University of Technology “KPI” Igor Sikorsky, ITS, Industrialny Lane 30, 03056 Kiev, Ukraine [email protected] 3 Lviv Polytechnic National University, ITRE, Professorska 2, 79013 Lviv, Ukraine [email protected]

Abstract. Cyber-PHY, IoT, sensor networks, Robotics (thick and server-less mobile applications), real-time network applications (thin clouds clients) can generate large arrays of unmanaged, weakly structured, and non-configured data of various types, known as “Big Data”. With the acceleration of industrial development “Industry 4.0” processing of such data became considerably more complicated. However, so-called problem “Big Data” is hard to solve or resist nowadays. The paper discusses the Best Practises and Case Studies for Data Analytics aimed to overcoming of the Big Data problematics under a slogan: “From Big Data to Smart Data!” Keywords: Big data · Industry 4.0 · 5G · IoT and robotics · Blockhain · Analytics and data mining · Ontology · Cloud and fog · Veracity

1 Motivation Big Data accumulation (Fig. 1) is nowadays typical for trading and marketing, electronic payments, production process, for the traffic from mobile providers, international justice and forensics, for public fiscal authorities, pharmaceutical and advertising industry. A large number of research institutes, organizations and universities accumulate, store and process large amounts of technical and scientific information. The sources can be primarily classified as follows [1–4]: • • • • •

Data from sensors and production capacities (Smart Manufacturing) Navigation data from GPS and other navigation systems, e.g. Galileo Data from GIS (Geographic Information Systems) Traffic and telematics data (IoT) Robotics data (navigation, sensing, measurements)

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 23–40, 2021. https://doi.org/10.1007/978-3-030-58359-0_2

24

A. Luntovskyy et al.

Fig. 1. Sources of unmanaged and unstructured dataunder Industry 4.0

• • • •

QoS management data and traffic data from mobile providers Power and renewable energy stations (data for smart metering) Data from scientific simulations and parallel computing (clusters, grids) Imaging data of patients from CT and MRT (Computer and Magnetic Resonance Tomography) in healthcare • Marketing information from cookies, social networks and Web2.0 semantic networks • Collected personal data of citizens and much more. In the conditions of modern industrial development, under so-called “Industry 4.0”, there are even more Big Data sources: home automation, patient health data, M2M and robotic data, business intelligence, pharmacological research, networking and experimental data. The mobile networks and aps for the 5G will definitely take an active part in the process of receiving and processing of large data amounts [2, 5–7] too. By year 2020, the new 5G networks will use more than 50–100 billion sensors to download comprehensive information about how we interact with things that surround us or even are inside of us?! As one of the most interesting further topics the Blockchain technology occurs. This is nowadays exponentially increasing and enables modern cryptocurrencies: e.g., Bitcoin, Monero, Ethereum etc. The technology of Blockchain and its associated applications like cryptocurrencies are so-called “resource eaters” due to their enormous energy and memory consumption. Large amounts of chained crypto blocks are causing surely “Big Data problematics”. Long-term mass data retrieving and processing requires large amounts of memory: decentralized, i.e. on private hosts and even on mobile devices, as well as centralized, in data centres and clouds. This means Big Data shortcoming too.

From Big Data to Smart Data: The Most Effective Approaches for Data Analytics

25

Another general problem is the economic and technological recognition of cryptotechnologies, processing of mass tests for the validity of blocks, the prevention of criminal activities (extortion, fraud) based on Blockchain [2, 18, 26]. The further paper content was built as follows: • State-of-the-art approaches are analysed (Sect. 2). • Big data shortcomings for important areas (best practices/case studies 1–5) are discussed like WSN, health care, IoT and robotics (Sects. 3, 4 and 5). • Analytics placement options to overcome Big Data are offered (cloud and fog based). • Regular paradigms for Big Data are overviewed (Sect. 5) as well as empirical Data Analytics for Big Data are discussed (Sects. 6, 7 and 8). • Finalising, conclusions and outlook have been done.

2 State-of-the-Art Huge as well as heterogeneous data volumes (approx. 100 PB to 100 EB) are nowadays so immobile that their access and management becomes only very efficient last time without use of special techniques [1–3]. Big Data are nowadays too complex for their efficient processing with classical manual methods for data structuring. Herewith a recognition of the problem is given by one of the world’s leading research and consulting company as very serious: the company Gartner Inc. notes “the problem of Big Data as one of the most important trends of IT-infrastructure development along with virtualization and energy efficiency of IT”. Under use of convenient retrieving, archiving and analytics methods for Big Data from [1–3] like databases and data warehouses, electronic sheets, formatted texts, CSV-data, Web- and XML-documents, graphical, audio and video documents data become faster suitable for nothing and inadequate. To solve this problem the computational techniques of so called “Data Mining” inter alia with ontologies, data pattern recognition, fuzzy logic etc. are used. These techniques can correspondingly be based on cloud analytics and disassemble the complexity and heterogeneity [1–13]: • • • • • • •

Taxonomies and predictive classification. Preliminary clusteringand compression. Predictive deviation detection (based on regular statistics). Descriptive association rule discovery (fuzzy based). Ontologies and knowledge data bases (KDB). Regression Models. Sequential Pattern Discovery (Machine Learning).

Otherwise, such a large amount of data remains unstructuredandwillbe in future characterized by the over-proportional management complexity. Additionally, significant increase of network traffic can be forecasted during the data processing under use of typical today solutions like Cloud Computing with heterogeneity of networks, replication across multiple geographically-distributed computing nodes [2, 6, 7, 10].

26

A. Luntovskyy et al.

3 The 6V Prevention Factors However, there are six factors, which affect the complication or prevention of the use of the regular and/or (cloud-based) analytical methods (Fig. 2). They build so-called 6V paradigm which distinguishes from known 5V or even 10V-paradigms [1, 3]: • Volume: the amount of generated and stored data is moving from 100 PB up to 100 EByte nowadays. • Velocity: it means the frequency and speed of data acquisition and processing (from slow batch techniques to real time with strong limitation on reaction time or latencies). • Variety: multiple codec types of multimedia data. This helps people who analyse it to effectively use the resulting insight. • Veracity: The quality of captured data can vary greatly, affecting accurate analysis. • Violation: The discussed data are often scattered, without any clustering or any structuring. • Value: Inconsistency of the data sets can negatively affect the processes of data handling and management. Unfortunately, the depicted V-factors for Big Data are growing faster than the performance of their analysis (in Mbyte/s or in GFLOPS) via classical computational techniques. Therefore, for Industry 4.0, IoT systems and M2M not only improvements in networking but also in analytic blocks became very significant [1–7, 27, 29], which aimed to decreasing of data streams and their algorithmic complexity.

Fig. 2. Own 6V paradigm

4 Unmanagedand Unstructured Data in “Industry 4.0” The goal of clustering and further processing of unmanaged and unstructured data from Industry 4.0 is the reducing of costs for their storage, better traceability, forecasting in process trends, workflow and production process optimization, data security and human private sphere protection. The paper discusses the Best Practises and Case Studies aimed to overcoming of the Big Data problematics if possible under use of “data compression” via their transformations (Fig. 3).

From Big Data to Smart Data: The Most Effective Approaches for Data Analytics

27

The “heavy-weighting approaches” involving precise analytical techniques and expensive specialized software are used for this aim. On the other hand, there is the opportunity to solve the Big Data problem under use of some “light-weighting approaches” based on agility: freeware, multipurpose techniques, minimal challenges on the personnel training and competencies. The paper examines the techniques and case studies of both topics. The “heavy-weighting approaches” (ontologies, knowledge bases, fuzzy logic and fuzzy knowledge bases) are compared to “light-weighting” one. The existing reference solutions are discussed below. So-called “Data Scientists” need to answer the following questions [8–16]: • How is the acquired data available? • When and with which periodicity the data survey acquisition and accumulation must be done? • In which way must data samplingbe done? • How can the data be accessed? • Which formats and codecs are required to apply within data evaluation system or for processing in cloud analytics blocks?

Fig. 3. Method classification: what is the aim of the clustering and further processing of unmanaged and unstructured data?

The typical sampling, survey and accumulation models are shown in Fig. 4. Mostly, the collected data is large unstructured and unmanaged.

28

A. Luntovskyy et al.

Fig. 4. Typical data acquisition and evaluation model

Example 1: Data Analytics for “Smart Data”. Depending on the event frequency, the event driven, periodically, permanent, and the behaviour of the sensors (push, pull), a sensor survey and data accumulation are performed. Let’s a WSN consists of 15000 sensors. Each sensor can transfer a short telegram up to 100 Bits. Thereby: • The survey for each sensor is conducted 20 times per hour: 2000 Bits/h • × 24 h = 48000 Bits/daily = 6000 Bytes/daily for each sensor • In general, an average sensor wireless network accumulates experimental data for 3 years × 365 days: • × 365 × 3 = 6.57 MB for each of the sensors • The overall-data for the mentioned network: • 6.57 MBytes × 15000 sensors = 100 GB of raw data! Further research of the data accumulated in this way can be carried out by manual and automated methods both (Fig. 5). As a result of such research data evaluation the following common documents can be issued [15, 16]: • • • • •

Management Template. Production Report. Error Report. Fault Forecast. Feedback Recognition.

From Big Data to Smart Data: The Most Effective Approaches for Data Analytics

29

Fig. 5. Data analytics: manual and automated

Often the data research cannot be completely studied or their trends recognized due to the complexity and heterogeneity. The future transition “Big Data – Smart Data” is given in Fig. 6.

Fig. 6. Transition “Big Data – Smart Data”

The overcoming of myths and misconceptions is possible due to use of Machine Learning (ML), refer Sect. 8.

5 Big Data Problematics for IoT and Robotics The number of robots that co-operate with a person, with machines and with each other in various fields of industry and economy, everyday life and the entertainment industry,

30

A. Luntovskyy et al.

social pedagogy is growing rapidly. The wide deployment of robots and robotics data create the right conditions nowadays for Big Data too. Up-to-date robotic platforms (Fig. 7) are able to use virtualization technologies and virtual robots over the network [2, 6, 17–22]. Example 2: Analytics Placement Options to overcome Big Data. The up-date IoT and robotic applications can be classified into three groups (Fig. 8). They solve the Big Data shortcomings in better way due to intelligent distribution of the Data Analytics [2, 6, 18, 22]: (1) Conventional Robots. (2) Cloud-Centric Robots. (3) Distributed (Fog-Cloud-cooperating) Robots. Only the cloud-centric solution (2) and further distributed, fog-cloud-cooperating solution (3) both are able to overcome the discussed problem nowadays and in the future. The analytic blocks, migration agents as well as further adaptive interfaces are delegated to the clouds and, possibly, after pre-processing and clustering backwards to the so-called “fog” (refer Fig. 8) under use of the mentioned solutions and protocols. The virtual analytical components (middleware, web services, and mobile software agents) are placed in the cloud and fog environment. Virtual cloud and fog solutions contain software components for the robots that implement reboot-able (virtual) business processes

Fig. 7. The properties of the future robotics platforms

From Big Data to Smart Data: The Most Effective Approaches for Data Analytics

31

[18–22]. Thus, we are talking about replaceable and customizable robotic algorithms (cp. Fig. 7 and 8) in various fields of application (industry, medicine, communication and telecommunication,entertainment) which can collect, process and retrieve voluminous heterogenic Big Data in the given area.

Fig. 8. Three communication ways: analytics placement options to overcome Big Data

6 Regular Paradigms for Big Data For the processing of Big Data the usual statistical paradigms can be deployed like S, R, SPSS, Oracle R, SAP Hana, IBM SPSS, Netezza, and Grafana [23–25]. R Language (1992) as well as R Project environment are aimed to the processing of significant statistic data (up to 106 samples), calculation of the statistic moment of the probability distribution as well as their visualisation under GNU license. In the integrated R environment there are the standard CLI and GUI. For data exports, graphic visualisation and report output all popular formats can be used: HTML, XML, LaTeX, PDF, MS Office. There are a lot of further R-editors and R-plugins, e.g. for Eclipse, Python, Gretl for all significant OS. SPSS (IBM 1968) is a trademark of IBM, which develops and distributes software for statistics and data analysis (IBM SPSS Statistics v25, PASW – Predictive Analysis SoftWare for MS Win, GNU/Linux, Mac OS X). R Language is interoperable to a lot of modern packages for Big Data processinglike SPSS, SAS, Platform Symphony, Tableau, SAP Hana, IBMNetezza. Recently, the open platform Grafana can be used for real-time visualizations of fast processes of inter alia network traffic data and real-time sensor data. The Grafana

32

A. Luntovskyy et al.

system is designed to construct dashboards and graphs and is a complex multi-level open source web application based on the Apache License. The Grafana web application supports, in particular, further server components such as Graphite, InfluxDB, Opentsdb (as the backend). Grafana is also suitable as an open platform for efficient analytics and monitoring of Big Data. The Big Data Appliance is NoSQL-Cluster from application servers for massiveparallel analysis based on the integrated R tools and Apache Hadoop.

7 Empirical Data Analytics: Case Studies/Best Practices Example 3: Ontology-Based Medical Imaging and Data Analytics. Medical formats for CT and MRT applications are based on the open standard DICOM (Digital Imaging and Communications in Medicine) founded by Siemens and Philips. DICOM stores, manages or transmits the compressed patient Big Data. DICOM standardizes both the data storage format and communication protocol, which is represented by digital medical images, data for their segmentation, surface definition or registration. The DICOM format is also the basis for archiving (Picture Archiving and Communication System, PACS) in the clinics and hospitals. The ontology-based framework for data analysis [3, 6] process enormous mass of patient data, which is stored in a variety of formats: handwritten texts, body imaging (CT, MRT), tests, forms with laboratory results, genomic data. Often such data should be processed in a few steps. Unstructured and unmanaged data must be converted to structured formats through their classification (taxonomies) and ontology creation The given method is a so-called “heavy-weighting” [3, 4, 8]. Example 4: A Light-Weighting Approach. Let us to examine a system example for overcoming of Big Data complexity which was developed via TIQ Solutions in Leipzig [3, 15, 16]. The mentioned reference solution of TIQ Leipzig is represented below. The system is based on “a light-weighting approach” and is oriented to processing of Big Data in the domain of Business Analytics. The system architecture is depicted in Fig. 9. The most important quality criteria of the referred solution are wide scalability and suitability for the Linux-clusters. The system allows the connectors to the conventional DB. The complex poly-structured and redundant retrieved data can be processed with higher performance within an enterprise or institution data centre or a cluster. Some in Java implemented modules allow even real-time control. The expenditures in the form of investments CAPEX (Capital Expenditures for hardware, cable infrastructure, premises) and OPEX (Operational Expenditures for licenses, personnel, electricity, ongoing maintenance) are significantly reduced on the basis of proven freeware components. Therefore, the solution is relevant for so-called middle-range enterprises. The discussed architecture is characterized viaagility, possesses no rigidity and offers only small license costsbytheuse of proven freeware. The components of the architecture [3, 15, 16] are as follows: • Apache Hadoop (2008) was implemented in Java and represents a freeware framework for scalable distributed applications under use of Google-similar algorithms and GFS (Google File System), which supports computing with significant data (Petabytesarea) on the computer clusters.

From Big Data to Smart Data: The Most Effective Approaches for Data Analytics

33

Fig. 9. Architecture example for Big Data processing

• Apache HBASE is a scalable DB for management of Big Data within the Hadoop clusters. • Apache Phoenix is a massively paralleled RDB-Engine with OLTP concept for Hadoop with Apache HBase as back-store. SQL queries are processed, then compiled into the series of HBase scans and are orchestrated to produce the regular JDBC results. • Apache Hive: a DBMS for Hadoop, aimed to data aggregation, queries and analysis with a SQL-similar interface. The datacanberetrieved in diverse DB and file systems. • Tableau (2003): software for data visualization and reporting. • Talend: cloud-based software for Big Data integration. • SCADA (Supervisory Control and Data Acquisition): the computer-assisted monitoring and control of technical processes.

8 Deployment of ML and AI Nowadays, artificial intelligence is widely used in all spheres of life without stopping in its development. In mid-term, the standards for ML and AI will accompany the industries, digital economy and everyday life over the world and for each institution. Surely, they will find it deployment in the area of Big Data. Therefore, let us to give a more detailed overview below. Frequently, to overcoming of misconceptions on Big Data the novel AI and ML techniques are used. The intelligent algorithms can provide clusterization and data preprocessing too.

34

A. Luntovskyy et al.

When talking about artificial intelligence (AI), we should also mention ML as an important basis. To begin with, we would like to represent what the difference between the convenient deterministic algorithms (cp. classical flowcharts, like SSADM or PAP by DIN 66001) and Machine Learning [1–3, 27–30]. Figure 10 shows that in classic algorithms we have Input Data and an algorithm that allows us to get results. In ML, we have Input and Output Data, which help us to get a neural network a learning algorithm and it will help in the future to make the neural network more powerful via training.

Fig. 10. Comparison of ML to deterministic algorithms

ML systems, based on cloud-concentrated knowledge (knowledge databases), create statistics and regressions on obtained voluminous experimental data in background mode. An artificial system is “learned” from samples and examples and can summarize them after the completion of the study and evaluation (training) phase. The ML system recognizes templates and trends in research data. Thus, the ML systems within modern distributed applications (Fig. 11) can also evaluate data on representativeness and compliance. Three mostly used types of ML algorithms are depicted below. The main purpose of machine learning (ML) is to look for learning algorithms, performed without human intervention, where the time and amount of data required for training is one of the most important indicators of performance. The main ML approaches are traditionally divided into Supervised Learning (SL), Unsupervised Learning (UL), and Reinforcement Learning (RL). The third approach is mostly important for the presented work. It means the mostly appropriate deployment chances for Big Data and eco-system (refer Fig. 11).

From Big Data to Smart Data: The Most Effective Approaches for Data Analytics

35

Fig. 11. Three mostly used types of ML algorithms

Supervised Learning (SL) is an ML method that takes training data (organized into an input vector (x) and a desired output value (y)) to develop a predictive model for the traffic classes. It is easy for the humans, e.g. to understand, where in some types of data we have video, music or pictures, but for a robot it becomes a more difficult task. So, we translate these data into a robot-understandable language, and indicate, were we have each kind of data. This will be our training data, that the algorithm memorizes and can already analyze the data flow. Knowing what type of data the user applies, we will give the algorithm the ability to classify them correctly (refer Fig. 12).

Fig. 12. Supervised learning for Big Data clusterization

By UL (Unsupervised Learning), we allow the robot to learn on its own manner without giving the correct answers to the problem we want to solve. The aim is as follows: without knowing what kind of data we have, the data could be divided via the algorithm into several classes (Fig. 13).

36

A. Luntovskyy et al.

Fig. 13. Unsupervised learning for Big Data clusterization

Reinforcement Learning (RL) algorithm try to understand how to correctly configure parameters to achieve a specific goal (Fig. 14). Many problems and sequential tasks cannot provide a clear answer to a problem, and when it performs poorly, RL has already proven effective in many real world applications such as robotics, stand-alone helicopters and drones, fixed network routing, automated industry and hazardous works etc.

Fig. 14. Reinforcement learning for Big Data clusterization

The (cloud-based) ML system enables artificial creation of knowledge from the obtained voluminous experimental data in background mode. An artificial system is “learned” from samples and examples and can summarize them after the completion of the study and evaluation (training) phase. The ML system recognizes templates and trends in research data. Thus, the ML (cloud) system can also evaluate data on representativeness and compliance. Evidently, at this time, we can propose only one way to improve this approach [1–3, 27–30]. We can use the cloud-centric solution, i.e. the stored behavioral data from offered knowledge bases, which provide real-time monitoring and learning. We want to introduce the stored behavioral data from offered knowledge bases. In this case, we will collect the data of each sensor or robots and send them to a common knowledge database. This will be an analogy as a common knowledge base for Tesla’s autopilot,

From Big Data to Smart Data: The Most Effective Approaches for Data Analytics

37

which can help us to make “studying models” for robots or sensors. Then our devices such as sensors can automatically predict the mass movement of data into specific places [30]. Example 5: An Approach of Big Data complexity overcoming for Telecom Operator. The telecom companies process huge multimedia data amounts. Unfortunately, such highly dynamic sphere is characterized nowadaysby an essential lag in the discussed problematics. Telecom data require involving of special approaches, methods, algorithms and tools for Big Data analytics and processing (Fig. 15): • call options (different parameters of customer calls) • customer profiles (personal customer data, duration of customer calls and other used services, billing records) • network data (different parameters, operations, infrastructure parameters and technical data) • data from financial reports, questionnaires, advertising, company plans, users’ applications etc. Therefore, the following data taxonomies and data scales should be considered: • • • • •

amount of records amount of calling (and called) customers diversity in calling behaviour amount of used services and services delivery options (characteristics) transmitted video-volumes.

But every analytic request uses only part of all these data. In this regard, there is a need to systematize and classify data from various sources. Ontological models are used as a problem solution [27–29]. This is the first step for “Big Data” overcoming. The second step is loading the data to execute the query via the GW (gateways) with available Analytics (Decision Blocks). This requires a dynamic workflow generation mechanism based on the ontology of data, knowledge and computational components. One of the approaches to designing such kind of mechanism is discussed in [9]. The third important step is fuzzy knowledge base (FKB) design for possibility to use not very hard computational process during the telecoms operation. The huge amount of statistics can be classified and clustered using the algorithms considered in [27]. But every cluster is only one fuzzy rule. The rule set is a FKB, but to check it correctness it’s possible using so-called meta-graphs. FKB correctness checking is discussed in [10]. The obtained FKB can be used as the basis for a decision block [29] for IoT or others telecoms platforms (refer Fig. 15).

38

A. Luntovskyy et al.

Fig. 15. Cloud architecture with a decision block: approach 1-2-3 [29]

9 Conclusions and Outlook The given work is dedicated to Big Data sources in such modern issues like Industry 4.0, Robotics, IoT, 5G, Blockchain and best practices for analytics. The paper discusses the Best Practises and Case Studies aimed to overcoming of the Big Data problematics if possible under use of “data compression” via their transformations: • Own 6 V-paradigm on Big Data distinguishing is formulated. • Primarily the state-of-the-art approaches are analysed. The mostly common techniques classification and overview is given. So-called “heavy-weighting” (ontologies, FKB, meta-graphs) and “light-weighting” approaches are examined. • The Big data shortcomings for important areas (examples 1-5) were discussed like WSN, healthcare, telecom companies and robotics. • Regular solution paradigms for Big Data are overviewed (e.g. R Tools, SPSS, Grafana). • The (cloud and fog based) analytics blocks placement options to overcome Big Data are offered. • Best Practices for Empirical Data Analytics for Big Data are discussed. This work can be qualified as a “work-in-progress”. The authors try to find new efficient method combinations to overcome the above mentioned 6 V-paradigm. Acknowledgement. Authors’ acknowledgements to BA Dresden, the PhD students and to the colleagues J. Spillner with ZHAW Zurich, J. Pejas, A. Cariov, W. Rogoza, A. Konys with ZUT Univ. of Technology Szczecin, W. Dargie with Dresden Univ. of Technology, D. Blankenberg with TIQ Solutions Leipzig, T. Maksymyuk with LNPU Lviv for technical support, inspiration and challenges by fulfilling of this work.

From Big Data to Smart Data: The Most Effective Approaches for Data Analytics

39

References 1. Ulema, M.: Big data and telecommunications telecom analytics. In: Tutorial at International IEEE Conference on BlackSeaCom-2016, Varna, Bulgaria (2016) 2. Luntovskyy, A., Spillner, J.: Architectural Transformations in Network Services and Distributed Systems: Service Vision. Case Studies. Springer, Heidelberg (2017). 344p. ISBN 9-783-6581-484-09 3. Globa, L., Svetsynska, I., Luntovskyy, A.: Case studies on big data. J. Theoret. Appl. Comput. Sci. JTACS, Polish Academy of Science, Gdansk, No. 2, 2016. ISSN 2299-2634, 10p 4. Konys, A., Rogoza, W.: Big data and ontologies. In: Talk at ACS International Conference 2016 in Mi˛edzyzdroje, October 2016. 3p 5. Luntovskyy, A.: Advanced software-technological approaches for mobile apps development. In: 14th Internation IEEE TCSET-2018 Conference, Lviv-Slavske, 2018, 6 p. IEEE Xplore: https://ieeexplore.ieee.org/document/8336168/. https://doi.org/10.1109/tcset.2018.8336168. Accessed 01 Jun 2020 6. Luntovskyy, A.: SLMA and novel software technologies for Industry 4.0. In: 21-st International Conference on ACS-2018, Szczecin-Mi˛edzyzdroje. In: Peja´s, J., El Fray, I., Hyla, T., Kacprzyk, J. (eds.) Advances in Soft and Hard Computing, Springer Int., 12p. (Part of the AISC book series, vol. 889 (2018). https://doi.org/10.1007/978-3-030-03314-9-16, ISBN 978-3-030-03313-2 7. Luntovskyy, A., Guetter, D., Klymash, M.: Up-to-date paradigms for distributed computing. In: International IEEE Conference AICT 2017, Lvyv, pp. 113–119 (IEEE Xplore). ISBN 978-1-5386-0638-4. https://doi.org/10.1109/aiact.2017.8020078 8. Kuiler, E.: From big data to knowledge: an ontological approach to big data analytics. Rev. Policy Res. 31(4), 311–318 (2014) 9. Globa, L., Novogrudska, R., Schill, A.: The approach to engineering tasks composition on knowledge portals. In: Proceedings of SPIE, vol. 10445 (2017). https://www.scopus.com/ record/display.uri?eid=2-s2.0-85058990971&origin=resultslist&sort=plf-f&src=s&sid=15a c89516aee1e66c12de1910. Accessed 01 Jun 2020 10. Globa, L., Ternovoy, M., Shtogrina, O., Kryvenko, O.: Based on force-directed algorithms method for metagraph visualization. Adv. Intell. Syst. Comput. 342, 359–369 (2015) 11. Russell, S., Norvig, P.: Artificial Intelligence: A Modern Approach. Prentice Hall, Upper Saddle River (2010) 12. Jones, M.: Artificial Intelligence: A Systems Approach, Hingham, Massachusetts, New Delhi, Infinity Sci. Press LLC, (2008) 13. Marr, B., Wiley, J.: Big Data: Using SMART Big Data, Analytics and Metrics to Make Better Decisions and Improve Performance, Sons Ltd. (2015) 14. IBH Reports: IBH Dresden Workshop on 23 March 2017 (in German) 15. Blankenberg, D.: Big Data in der Industrie 4.0, TIQ Solutions Leipzig, Workshop an der IBH Dresden, 13 March 2018. (in German) 16. TIQ Solutions Leipzig: https://www.tiq-solutions.de/. Accessed 01 Jun 2020 17. Luntovskyy, A., Guetter, D.: Allgegenwaertige vernetzung: Industrie 4.0, Internet der Dinge, Fog Computing u.v.m! in BA SaxonyMagazine “Wissen im Markt”, Vol. 2/2018, Glauchau (2018). 8p. (in German) 18. Luntovskyy, A., Klymash, M.: Software Technologies for Mobile Apps, Apps for Fog Computing, Robotics and Cryptoapps, Lviv (2019). 247 p. (Monograph, in Ukrainian) 19. Luntovskyy, A.: Modern apps and platforms for the robots. In: International Conference on Theoretical and Applied Aspects of Modern Infocommunications, Vyshkiv-Carpathians2019, January 2019. 4p

40

A. Luntovskyy et al.

20. Spillner, J.: Serverless Cyber-Physical Applications, Keynote Talk at Workshop “Science meets Industry. Fraunhofer-Institute Centre Dresden (2018). 20 p 21. Luntovskyy, A., et al.: Intelligent networking and bio-inspired engineering. In: International IEEE Conference on UkrMiCo-2017, Odessa, Ukraine (2017). 4p. (IEEEXplore: http://iee explore.ieee.org/document/8095421/). Accessed 01 Jun 2020 22. Int. Federation of Robotics. http://www.ifr.org/. Accessed 01 June 2020 23. R-Project. http://www.R-project.org/. Accessed 01 June 2020 24. SPSS: http://www.ibm.com/analytics/de/de/technology/spss/. Accessed 01 Jun 2020 25. Grafana. http://docs.grafana.org/. Accessed 01 Jun 2020 26. Luntovskyy, A., Guetter, D.: Cryptographic Technology Blockchain and its Applications, Proc. of UkrMico-2018, Odessa, Ukraine, Springer Int. (2019). 15 p 27. Globa, L., Novogrudska, R., Koval, A.: Ontology model of telecom operator big data. In: International IEEE Conference BlackSeaCom 2018. https://ieeexplore.ieee.org/xpl/mostRe centIssue.jsp?punumber=8419227. Accessed 01 Jun 2020 28. Globa, L., Kurdecha, V., Ishchenko, I., Zakharchuk, A.: Fuzzy logic usage for the data processing in the IoT Networks. In: Proceedings of OSTIS 2018, Minsk, Belarus, pp. 287–292 (2018) 29. Globa, L., Kurdecha, V., Ishchenko, I., Zakharchuk, A., Kunieva, N.: The intellectual IoTsystem for monitoring the base station OoS. International IEEE Conference on BlackSeaCom 2018. https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8419227. Accessed 01 June 2020 30. Maksymyuk, T., Han, L., Larionov, S., Shubyn, B., Luntovskyy, A., Klymash, M.: Intelligent spectrum management in 5G mobile networks based on recurrent neural networks. In: IEEE CADSM 2019, IEEE Xplore: https://ieeexplore.ieee.org/document/877930. Accessed 01 Jun 2020

Cloud-Based Architecture Development to Share Vehicle and Traffic Information for Industry 4.0 Serhat Bulut Ibrahim(B)

and Haci Ilhan(B)

Department of Electronics and Communication Engineering, Yildiz Technical University, Esenler, 34220 Istanbul, Turkey [email protected], [email protected]

Abstract. The automotive and technology sectors have developed very rapidly over the past decade. In addition to this growth, it also introduced the term Industry 4.0, which is used to represent the current Industrial Revolution. This revolution encompasses many sectors from manufacturing to health care. With Industry 4.0, digital transformation can create value throughout the entire product lifecycle, support customer feedback, and provide advanced solutions to the problems to be experienced. Automatic communication between the vehicle and the management office will facilitate our lives by enabling different analysis of vehicles such as the driver’s vehicle usage history, fuel consumption, maintenance indicators, determination of a behavioral model with data like temperature. This provides analysis of the car, the driver’s experience, and the road, preventing critical problems and unwanted behavior, and increasing safety on the roads. For example, during the winter season, municipal employees will not have to wait at night to intervene in road freezing. Employees will instantly monitor which roads are at risk of icing through the data collected from the vehicles on the road and municipal workers will salinize them in no time. This article aims to implement a platform that collects and analyzes vehicle sensor data and provides individual and corporate feedback. Using the OBD-II scanner, it is intended to help prevent problems, reduce accident rates and manage different types of vehicles. A ready OBD-II (On Board Diagnostic) reader device, supported by a Bluetooth connection, is used to collect data directly from the ECU (Engine Control Unit) in real-time, using an androidbased smartphone as a cloud network connection. The architectural structure in the cloud is capable of collecting and analyzing raw data to detect the occurrence of errors in vehicles. While providing feedback to the user, a smart cloud-based architecture provides the necessary information to the relevant municipal, fire or ambulance units by foreseeing traffic accidents, the icing on roads in winter or asphalt melting in summer. Experiments and tests conducted in Istanbul’s traffic show that the proposed platform has applicability and potential to use. Keywords: IoT · Industry 4.0 · Vehicle networks · Intelligent Vehicle · Data collection · Data processing · Data management · Cloud-based software architecture

H. Ilhan—Senior Member, IEEE. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 41–54, 2021. https://doi.org/10.1007/978-3-030-58359-0_3

42

S. Bulut Ibrahim and H. Ilhan

1 Introduction The number of vehicles is increasing day by day. According to the increase rate, it is predicted to reach 2 billion in 2035 [1]. Most of these vehicles will communicate with various types of devices and equipment via wireless communication protocols and broadcast media [2]. In addition, it is envisaged that by the year 2020, 25 billion devices will have been connected to the Internet through the Internet of Things [2]. In this context, it has been found that the raw data obtained from these devices and vehicle systems can be automatically received by the vehicle. All data can be used for interactions in a new processing and communication context by changing the direction of vehicles are used. Such interactions take place from the perspective of the Internet of Things (IoT), which enables the exchange of information between objects and plateaus [3]. The resulting paradigm encompasses the infrastructure of hardware, software, and services that connect physical objects called “Internet of Things” [4]. This scenario allows the definition of IoIV (Internet of Intelligent Vehicles) [5], which aims to provide a range of new services, such as fleet management systems, improving vehicle safety, efficient energy/fuel consumption, traffic planning, and accident reduction. This issue will gain even more importance with autonomous vehicles [6]. IoIV and Industry 4.0 have the potential to create a new paradigm of systems [7], providing real-time customer feedback between end-users (vehicle owners) and related organizations. Thus, it will be possible to make a change in a number of products-systems and transfer all data to the cloud in order to respond more quickly to customer requirements as shown in Fig. 1.

Fig. 1. Access to vehicle and user data via the cloud

This study aims to develop customer feedback and information sharing platform for Intelligent Vehicles in the perspective of digital transformation. A ready OBD-II (On Board Diagnostic) reader device, supported by a Bluetooth connection, is used to collect data directly from the ECU (Engine Control Unit) in real-time, using an android-based smartphone as a cloud network connection. The architectural structure in the cloud is capable of collecting and analyzing raw data to detect the occurrence of errors in vehicles.

Cloud-Based Architecture Development to Share Vehicle and Traffic Information

43

2 Materials and Methods 2.1 General Architecture of the System The architectural structure developed with this article consists of three major main modules. These are; 1. OBDII system where vehicle sensor data is read from EM327 socket, 2. A mobile application transmitting vehicle data, smartphone sensor data to the Cloud and showing interpreted data and information, 3. The decision support system in the cloud where all data will be collected (Fig. 2).

Fig. 2. Connecting all vehicles to the cloud via smart mobile phone

In the scope of the article, data obtained from both the driver’s smartphone and the sensor information in the vehicle are provided as an architectural infrastructure to be interpreted using artificial intelligence, big data analysis, and machine learning techniques and make various suggestions to the user. Interpretation of this data is performed in the cloud environment, and it is planned to contribute to the other drivers in the traffic using the system with information and shares. 2.2 Equipment and System Vehicle Connection Module (OBDII) Vehicle connection is achieved by OBD-II technology, which provides a real-time communication interface between sensors and triggers present in the vehicle.

44

S. Bulut Ibrahim and H. Ilhan

The interface connector of the OBD-II, known as the Data Link Connector (DLC), has been made mandatory in the USA since 1996, in the European Union countries since 2003 and all vehicles have been manufactured in Brazil, China, and Russia since 2010. OBDII connects to the vehicle’s brain and uses Bluetooth or wifi protocols to transfer data on the vehicle. For example, communication protocols: CAN, SAE J1850 PWM-VPW, and ISO 14230 KWP2000) [10]. In addition, OBD-II technology supports ten operating modes. Each has a specific set of commands that return data from the sensors and the computer in the vehicle. To request such data, codes called “Parameter Definition” (PID) are used. However, car manufacturers do not have to support all modes of operation in their vehicles. Therefore, finding a procedure is necessary to decode all PIDs supported by a tool. Table 1 provides the corresponding operating modes. Table 1. Operating modes supported by OBD-II Mode

Explanations

01

Returns real-time ECU data

02

Requests ECU data corresponding to final failure

03

Displays error codes stored in the vehicle

04

Clears stored error codes

05

Returns test results of O2 sensors available on the vehicle

06

Returns test results which are not related to continuous monitoring

07

Returns test results which are related to continuous monitoring

08

For control of built-in systems

09

Returns vehicle information

10

Shows error codes with persistent status

The OBDII device used in our Renault Megane test vehicle is as depicted in Fig. 3.

Fig. 3. OBDII scanner which we used in our project

Cloud-Based Architecture Development to Share Vehicle and Traffic Information

45

The socket on which the OBDII is installed in the test vehicle is shown in Fig. 4.

Fig. 4. Example of EM327 socket in car

Android is a mobile operating system developed by Google. It is based on a modified version of the Linux core and other open-source software and primarily designed for mobile devices like smart touch phones and tablets. Also, Google developed Android TV for televisions, each having a private user interface, Android Wear OS for cars, and Wear OS for wristwatches. Android has been the world’s best-selling operating system on tablets since 2011 and smartphones since 2013. As of 2018, it has more than 2.6 million apps in Google Store [10]. Table 2. Mobile operating system market rates (February 2019).

80 70 60 50 40 30 20 10 0

70.76

23.82 5.42 Android

iOS

Others

According to Statcounter’s report [11] published in February 2019, the Android operating system among the operating systems of smart mobile phones and tablets across the globe was preferred with 70.76%. In Table 2, it is presented in detail. Since Android is the most preferred operating system and open-source code in the world, we chose to write code to the Android operating system.

46

S. Bulut Ibrahim and H. Ilhan

Cloud is a term that usually refers to access to data centers via a wide area network (Wireless Ad-Hoc Networks, WAN) or an Internet connection, to a computer, information technology (IT), and access to software applications. Almost all information technology resources can take part in cloud technology and can be used via it. (Such as a software program or application, service or an entire infrastructure.) For instance, if an enterprise wants to build an IT infrastructure, it usually installs the servers, software, and network resources it needs, but almost all of these services and resources are accessible by going to third parties in the cloud. Cloud computing has many advantages. This service can be instantly accessed in most cases. Remote users can access cloud resources from anywhere they are connected to, rather than being limited to physical geography. It is usually divided into three categories: individual, showing who has access to services or infrastructure, general and mixed. Public cloud services are available to anyone who wants to buy or rent services. Businesses establish private cloud services only for use by employees and partners. Hybrid cloud services combine both. In this study, Hybrid Cloud is emphasized in the recommended system. 2.3 Recommended System Architecture Output Transferring Sensor Data from Vehicle to Cloud System: For example, transfer of the data of a vehicle with a temperature of 34° and going from Ata¸sehir at a speed of 89 km, to the Cloud system running at 2.3 rpm is carried out with the following steps. Obtaining Vehicle Data with OBDII Device: After the OBDII device is connected to the EM327 socket at the bottom of the test instrument console, the LED light on the back starts blinking and it is detected by smartphones with Bluetooth capability. Connecting an OBDII Device to a Smart Android-Based Mobile Phone: After activating the Bluetooth feature of the Android-based smartphone, devices with Bluetooth feature in the surrounding are listed on the screen of the mobile phone. OBDII device is selected from this list and the password is entered to initiate the communication between the OBDII device and the smartphone. In the car; The data on sensors such as petrol, wiper, battery, parking, speed, pedal, etc. are started to be read from the EM327 socket for operation with Bluetooth-enabled OBDII device (Fig. 5).

Fig. 5. Communication between the car and the smartphone

Cloud-Based Architecture Development to Share Vehicle and Traffic Information

47

Transferring Data from the OBDII Device with a Post-connection Application to the Cloud: With the developed android application, data such as OBDII device, the speed of the car, engine speed, engine load, hex number expression are taken and the package is sent to the cloud by being periodically packaged. (The codes for different brands and models of vehicles are different from OBDII.) Sending the Collected Data to a Cloud via Smartphone

Fig. 6. Communication of smart mobile phones with the cloud

Collected data is converted to key values and in the Json (JavaScript Object Notation) format, the https protocol and the post method are sent as packets to the cloud environment in a separate communication with the processing script. An example of supported Json format follows as (Fig. 6, Tables 3 and 4): Table 3. Attribute keys Attribute key Type Account

accountDataObject

Vehicle_data List[metricObject] Fault_codes

List[faultCodesObject]

Phone_codes List[metricObject]

Vehicle sensor data from the Android-based smartphone to the cloud environment passes through the main layers: Data Collection (Collection Layer), Processing Layer, Data Visualization (Data Visualization Layer), Application Layer, Notification Layer. The data of these layers is to transfer the tasks to the next subdivisions in detail.

48

S. Bulut Ibrahim and H. Ilhan Table 4. Example of JSon model

Json models

Fields

Types

Examples

accountDataObject

timestamp

String

2018-03-24T11:56:34+03:00

transactionId

String

1AB23456C***

clientIP

String

127.0.0.1

version

String

0.1.0

userId

Integer

1

cardId

Integer

1

phoneId

Integer

1

timestamp

String

2018-03-24T11:56:34+03:00

code

String

speed, eng_rpm,

value

Integer

48

timestamp

String

2018-03-24T11:56:34+03:00

PID

Integer

4

DataLen

Integer

1

Desc

String

Calculated engine load value

metricObject

faultCodesObject

3 The Architecture of Cloud Side A. Collection Layer As an example, when a user and a car are added to the system consisting of individual numbers (ides) on the cloud side for all systems, users, cars, mobile phones, these numbers are sent to the cloud system in each Json message as userid, carid (cardid) and mobile phone number (phoneid). In addition, in order to prevent system security, user (username), password and user information are also transferred to the cloud system in

Fig. 7. Cloud-based and artificial intelligence supported information sharing system architecture

Cloud-Based Architecture Development to Share Vehicle and Traffic Information

49

Json format. Data from mobile phones is sent in Json format in a single batch at certain periods. (If needed, emergency data can be sent (Fig. 7).) For an application developed on an Android-based smartphone, a version of each software update will exist in the application market. For example, if version 0.1.0 is installed on the user’s smartphone, the data from this app is indicated in the app version sent in the Json format to the application running at https://endpoint.app/v1/. If a new version of 0.2.1 is installed for use on smartphones after a certain period, this version sends to https://endpoint.app/v2/. (It is sent as Json version 0.2.1.) The collected data is transferred to Fluentd, an intermediate storage layer. Data from the cloud in Json format is written to Apache Kafka using Fluentd data collector. In order to be able to collect big data blocks quickly and accurately and transfer them to other systems, we needed a messaging system (queue). At this point, Apache Kafka assigns the flowing data into a queue, allows us to transfer to the Elasticsearch system. In addition, our reason for storing Apache Kafka storage reason is to prevent data from coming under heavy load and the system to be affected by instant densities as well as preventing data loss in future systematic operations (such as system update). Load Balancer is put to distribute the load in front of the apices in the cloud and to enable the different versions installed on the smart devices in the field to be routed to specific API. When each GooglePlay or Apple Store new version application is existed, since smartphones with older versions cannot perform a version update at a time, while V1 looks at the previous APIcluster, the V2 looks at the next more recent cluster. New published APIs, again according to their version, write separate Kafka message queue data (message queue). Data from users’ smart android mobile phones to the cloud are written on different topics such as Collectv1_1 and collectv2_1. These APIs direct the incoming data directly to the Kafka message queue without any action. Apache Storm is used as the event processing topology. In this way, vehicle and vehicle sensors from OBDII were collected in the cloud. B. Data Processing Layer The data has been received by pulling it from Apache Kafka, the hash code from this stage will be calculated and the data will be converted to events (if multiple events are required). Later, this data will be enriched by combining the data held on the memory. Many users and system-based rules will be enriched by questioning the event type, such as car models, defined speed rules, etc. Enriched events will be passed through a sophisticated event processing engine and new events to be produced will pass through the same cycle. Finally, the events will be sent to the storage layer for storage. Also, some of the data will be directed to the data visualization layer for car enriched with events and dashboard displays that reflect user states. For example, there is a process as follows. 1. On mobile phones, the application is installed in different versions, and the incoming data varies. The data from Collectv1_1 and Collectv2_1 passes through preprocessing, so all data is translated into common format.

50

S. Bulut Ibrahim and H. Ilhan

2. Enrichment is used to enrich the data using the external cache hazel cast for incoming data. Using the user, tool and application numbers (ides) in Json format, car information is taken. Considering the personal data security clauses in the European Union security laws, the brand model that comes to the cloud is combined with the information in the user information database, the user number is also found and enriched with emphasis on user privacy. Then the user adds the tool on which the user is installing the OBDII scanner. This car information is also being recorded on the application database (APP DB). After each car and phone user registration Data on the APP DB is automatically loaded into the Hazel cast, e-cache or, if any data has already been added, the data in the cache is updated. For example, if the user’s unicorn exceeds 80 km, this information is added to both the application database and the hazel cast if it adds a notification rule. 3. Rule processing is performed on flow data. Incoming data becomes a table. Notification messages become a speed table. 4. If the rule matches, SMS, mail, or instant notification (push notification) are sent according to the preference. For example, if the notification is set to 80 km, the following selection of a database selection statement works. select * from speed(1 m) where user_id = id and 80 < avg(speed). 5. The incoming events are stored in the Cassandra database after the Json is enriched.

C. Data Visualization Layer Based on this layer, Elasticsearch is used which is nosql db with search engine technology. In order to quickly process events, Elasticsearch was preferred. Through web applications and mobile applications with Angular2 and Elasticsearch dashboards are offered to users. D. Application Layer The application database will be located in this layer; as a database, myslql is currently being used (mariadb or postgresql migration is planned). On this db customer will keep the car campaign information and rules, notification settings. Rest APIs, which will be used by mobile application and web interfaces, are being developed on the web app. on this layer. (For example rest APIs, etc.) E. Notification Layer Rule operations are performed on the live data. Installed rules are placed in the queue, which phone, which e-mail address should be sent by type (SMS, mail, instant notification). F. Data Storage Two different databases have been created: Cassandra and mysql to which web servers are connected. The main reason for creating two databases is to maximize performance, and it will provide a smooth transition when the capacity is increased in the future. Many large companies such as Facebook, Twitter, Netflix use Apache Cassandra. For example, the smartcar.deviceinfo table in which the sensor data from the OBDII is transmitted to the cloud via the smart mobile phone is shown in Fig. 8.

Cloud-Based Architecture Development to Share Vehicle and Traffic Information

51

Fig. 8. Vehicle sensor data from OBDII

4 Results The data held in the database is presented to the user through the developed interfaces. With the coordinate information held in the database, the map shows where the vehicle is located using Google Maps in API. For displaying data from the OBDII device on-line, when the corresponding tool is selected, the enriched data, as shown in Fig. 9, is listed on the right side of the screen. Since we conducted our tests in Istanbul in May and June, the measured values were generally between 20° and 30°. If the temperature data from outside the motor is measured during the winter months, it can be determined that the temperature values at the relevant locations along with the coordinate information proceed in the negative direction. It is determined where there is a risk of icing in the city. Municipal officials can also perform salting tasks quickly. As the temperature graph goes above 40° in the summer, asphalt will melt, they can also look at the system and see which roads and coordinates the asphalt can melt. In this way, we do not risk the lives of municipal officials guarding the streets at night during the winter months.

52

S. Bulut Ibrahim and H. Ilhan

Fig. 9. Vehicle location and sensor data listing screen

The temperature data from the temperature sensor entering the engine from the outside along the route of our vehicle is shown in Fig. 10.

26.9

27 26.8 26.5

26.6 26.2 26.1 26.1

26.3826.4

26.3

26.3

26.4 26

26.8

26.1

26

25.8 25.6 25.4 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Fig. 10. Air temperature measured along the route

Cloud-Based Architecture Development to Share Vehicle and Traffic Information

53

The velocity graph is measured on the same road and at the same time as in Fig. 11.

60

56 51

50

40 42

40 30

45 38

41

44

33 25

22

20 10 0 0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

Fig. 11. Speed of test vehicle measured along the route

5 Conclusion In this article, an architectural structure proposal and pilot study for Industry 4.0, aimed at collecting and analyzing data from OBD-II hardware, from all sensors in vehicles, are presented. It has been seen that the proposed architecture can benefit research and development in the region. 4.5G technology is used for data transfer to the Cloud via mobile phone and tested on a single-vehicle. Our goal in the next step is to work with more vehicles and make improvements by controlling how the system behaves. As the data in the system increases, more meaningful data will be obtained by inserting big data into artificial intelligence. Besides, there were problems with software performance during platform development. Web service response times have been reduced from 8 s to 0.9 s. By bringing together many current systems, the cloud-based decision support platform, which works successfully from end to end, has been designed to provide information to drivers, institutions and organizations in traffic, minimizing their mistakes and wrong decisions and thus reducing traffic accidents and road conditions.

References 1. Yang, F., Wang, S., Li, J., Liu, Z., Sun, Q.: An overview of internet of vehicles. China Commun. 11, 1–15 (2014) 2. Contreras-Castillo, J., Zeadally, S., Ibáñez, J.A.G.: A seven-layered model architecture for the internet of vehicles. J. Inf. Telecommun. 1(1), 4–22 (2017)

54

S. Bulut Ibrahim and H. Ilhan

3. Gubbi, J., Buyya, R., Marusic, S., Palaniswami, M.: Internet of Things (IoT): a vision, architectural elements, and future directions. Future Gener. Comput. Syst. 29(7), 1645–1660 (2013) 4. Zanella, A., Bui, N., Castellani, A., Vangelista, L., Zorzi, M.: Internet of things for smart cities. IEEE Internet Things J. 1(1), 22–32 (2014) 5. Dandala, T.T., Krishnamurthy, V., Alwan, R.: Internet of Vehicles (IoV) for traffic management. In: 2017 International Conference on Computer, Communication and Signal Processing (ICCCSP), pp. 1–4 (2017) 6. Sugayama, R., Negrelli, E.: Connected vehicle on the way of Industry 4.0 (2017) 7. Lin, D., Lee, C., Lau, H., Yang, Y.: Strategic response to Industry 4.0: an empirical investigation on the Chinese automotive industry. Ind. Manag. Data Syst. (2018) 8. Pieroni, A., Scarpato, N., Brilli, M.: Industry 4.0 revolution in an autonomous and connected vehicle a non- conventional approach to managing big data. J. Theor. Appl. Inform. Technol. 96(1) (2018) 9. Hamidi, S.R., Ibrahim, E.N.M., Rahman, M.F.B.A., Shuhidan, S.M.: Industry 4.0 urban mobility: goNpark smart parking tracking module. In: Proceedings of the 3rd International Conference on Communication and Information Processing, ICCIP 2017, pp. 503–507. ACM, New York (2017) 10. Baekand, S.H., Jang, J.W.: Implementation of integrated OBD-II connector with an external network. Inf. Syst. 50, 69–75 (2015) 11. Number of available applications in the Google Play Store from December 2009 to December 2018. Statista. Accessed 26 Jan 2019 12. https://gs.statcounter.com/os-market-share/mobile-tablet/worldwide/#monthly-201702-201 902-bar

Complex Approach in Cryptanalysis of Internet of Things (IoT) Using Blockchain Technology and Lattice-Based Cryptosystem Lela Mirtskhulava1(B) , Larysa Globa2(B) , Nana Gulua3(B) and Nugzar Meshveliani3(B)

,

1 Iv. Javakgishvili Tbilisi State University/San Diego State University, 0186 Tbilisi, Georgia

[email protected] 2 National Technical University of Ukraine, Kiev, Ukraine

[email protected] 3 Sokhumi State University, Tbilisi, Georgia

[email protected], [email protected]

Abstract. In the given paper, we offer complex approach to secure The Internet of Things (IoT). IoT turns each physical object into a smart object sensing the environment surrounding us and brings new security risks in business sector, healthcare and in all the aspects of human life. The exponentially growing numbers of devices connected to IoT networks makes very hard to authenticate and secure all these devices within IoT network. It’s very difficult to maintain and manage a centralised IoT security model which is prone to failure and DDoS attacks. Wi-Fi technology is one of the leading technology. Wi-Fi uses radio waves prone to eavesdropping. WPA2 protocol security issues discovered recently by Computer Scientists and WPA3 has been developed to improve Wi-Fi security aspects but in a new WPA3-Personal protocol have been found the vulnerabilities allowing intruders to crack Wi-Fi passwords and intercept encrypted traffic sent between the Wi-Fi users. Blockchain mechanisms (BCMs) play a role in securing many IoT-oriented applications by becoming part of a security mosaic, in the context of a defenses-in-depth/Castle Approach. A blockchain is a database that stores all processed transactions – or data – in chronological order, in a set of computer memories that are tamperproof to adversaries. These transactions are then shared by all participating users. Information is stored and/or published as a public ledger that is infeasible to modify; every user or node in the system retains the same ledger as all other users or nodes in the network. This paper highlights some IoT environments where BCMs play an important role, while at the same time pointing out that BCMs are only part of the IoT Security (IoTSec) solution. We highlight blockchain’s important role for securing IoT. Cryptography technologies are the best solution to withstand communication attacks. We analyse NTRU cryptosystem capable to implement the NTRUEncrypt public key encryption algorithm withstanding the attacks from the quantum computers. Keywords: IoT security · Blockchain · NTRU cryptosystem · Encryption · Decryption © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 55–66, 2021. https://doi.org/10.1007/978-3-030-58359-0_4

56

L. Mirtskhulava et al.

1 Introduction The Internet of Things (IoT) makes our lives more comfortable but brings new security risks at the same time. IoT turns each physical object into a smart object capable to sense the environment surrounding us. These security issues can cause negative effects on various aspects of our lives at home, in the car we’re driving, and the effects that can cause our health issues. IoT security risks are extremely serious in business sector and In all the aspects of human life driven by information. It has become imperative to protect useful information from attacks. An attacker can hack the smart objects used in the big enterprises. Such cyberattacks are always addressed to steal very sensitive information about the company earnings and to intercept credit card information and etc. One of the largest hacking case took place in 2013 in US, where five hackers stole $160 million from credit cards [1]. IoT started emerging faster into the mainstream industry that can cause very serious security challenges for IoT system. More and more devices are connecting to the system with exponentially growing numbers what makes very hard to authenticate and secure all these devices in IoT network. It’s very difficult to maintain and manage a centralised IoT security model which is prone to failure and DDoS attacks. Wi-Fi technology is one of the leading technology in IoT systems and plays a very significant role but main challenge in using Wi-Fi networks is security issues. Wi-Fi uses radio waves prone to eavesdropping. WPA2 protocol security issues discovered recently by Computer Scientists and WPA3 has been developed to improve Wi-Fi security aspects. WPA2 uses a cryptographic - four-way handshake process for validation of the users. Main weakness in WPA2 was found in 2017 using “key reinstallation attacks” (KRACKs) [2]. In a new WPA3-Personal protocol have been found the vulnerabilities allowing intruders to crack Wi-Fi passwords and intercept encrypted traffic sent between the Wi-Fi users.

2 IoT and Its Security Issues and Capabilities The definition of the Internet of Things (IoT) by ITU-T - “A global infrastructure for the information society, enabling advanced services by interconnecting (physical and virtual) things based on existing and evolving interoperable information and communication technologies” [1]. IoT reference model (Fig. 1) consists of four layers plus security capabilities and management Capabilities. Four layers of IoT are: 1) application layer, 2) service support and application support layer, 3) network layer and 4) device layer. We focus on IoT Security capabilities such as generic security capabilities and specific security capabilities. Generic Security Capabilities Include: a) Authentication, authorization, privacy protection, security audit and anti-virus, Application data confidentiality and integrity protection at the application layer; b) Authentication, authorization, use data and signalling data confidentiality, and signalling integrity protection at the network layer;

Complex Approach in Cryptanalysis of Internet of Things (IoT)

57

c) Authentication, authorization, access control, device integrity validation, data confidentiality and integrity protection at the device layer; d) Specific security capabilities are closely coupled with application-specific requirements.

Fig. 1. IoT reference model. (Rec. ITU-T Y.2060 (06/2012))

Securing IoT is the biggest challenge for technology companies. In IoT network, data is collected from external sensors that are placed in public sites allowing anyone to send harmful data to the network. Bring your own device (BYOD) is another case when third-party devices are allowed to access the network [2]. There are most vulnerable areas of IoT given in the Table 1: Table 1. IoT vulnerabilities IoT security requirements

Description

Confidentiality

Ensures that the exchanged messages can be accessed only by the intended users

Integrity

Ensures that the exchanged messages were not modified by the intruder

Authentication

Ensures that the sender and the receiver involved in any operation are right identities avoiding a masquerade attack usually targeting this requirement and claiming to be another identity

Availability

Ensures that the service is not denied avoiding Denial of service attacks targeting this requirement and causing service disruption (continued)

58

L. Mirtskhulava et al. Table 1. (continued)

IoT security requirements

Description

Authorization

Ensures that entities are permitted to do the operation they request to perform

Freshness

Ensures freshness of the data

Non-repudiation

Ensures that an entity is unable to deny an action that it has performed

Forward secrecy

Ensures that after an object leaves the network, it no longer have an access to data exchange process

Backward secrecy

Backward Secrecy: ensures that a new object joining the network was unable to access the previous communications

3 Blockchain Enables a New Level of Trust in IoT Security Blockchain is a shared, immutable ledger that facilitates the process of recording transactions. The blockchain like a permanent book of records can keep a log of all transactions that have taken place in chronological order with timestamps in each block. Blockchain stores transaction data in blocks that are linked together forming a chain. Blocks can record and then confirm the time and sequence of the transactions logged into the blockchain within a network that may to be governed and agreed by the network participants following the specific rules. A block contains transaction data and other important details but it will always contain a timestamp, a unique hash and ma previous hash what makes blockchain more secure (Figs. 2 and 3).

Fig. 2. Blockchain structure

Complex Approach in Cryptanalysis of Internet of Things (IoT)

59

Fig. 3. Block structure

Properties in a Block: Timestamp: The time that determines when the block was created and its location on the blockchain. Transaction Data: The information to be securely stored in the block. Hash: The unique code generated by combining the contents for each block itself well known as a digital fingerprint. Previous Hash: Each block contains the hash of the previous block what makes the blockchain unique because this chain will be broken if a block is modified or decrypted. Hashing: This is an application of cryptography that is fundamental to the design of the blockchain. Cryptography it is a way to generate a random but the calculated string of number and letters from any kind of input. This is accomplished by the use of a hash function. A hash function like a machine can takes an object, such as an apple, and turn it into a combination of letters and numbers varying with the length and content like “1a432bf”. Gathering Blockchain Transactions: the transactions are carried out and they are placed in a special location called the mempool capable to collect all these unvalidated transactions. Very latest transactions stored in the mempool are broadcasted to all participants within blockchain (Fig. 4).

Fig. 4. Timestamp in hashing

60

L. Mirtskhulava et al.

Cisco Blockchain Framework Cisco offers the Blockchain framework enabling the parties to reach mutual agreement on the transaction authenticity. Blockchain is a decentralized technology enabling all the outcomes been permanently recorded throughout a shared database that is cryptographically secured. There are multiple blockchain configurations used to reach an agreement in blockchain network. Many enterprises are using a permissioned blockchain technology where only known objects may participate providing more privacy, highest speed, and better administer the network [4]. The autonomous and decentralized model of the blockchain is an ideal solution for the Internet of Things eliminating the failures by creating a more secure platform enabling the devices run on it. Cisco hardware-independent blockchain framework specifies a set of standards addressing infrastructure-level security risks based on reference architectures deploying on the premises: in an demilitarized zone (DMZ) of an enterprises, and in the cloud or a hybrid deployment. The proposed platform supports hardware security modules (HSMs) and WAN optimization tools such as software-defined WAN (Fig. 5).

Fig. 5. Cisco blockchain framework: infrastructure and network [4].

The distributed nature of the blockchain enables it to be inherently secure but without the correct blockchain design they it can be prone to the threats like distributed-denialof-service (DDoS); a collusion between blockchain nodes; the vulnerabilities in smart contracts including multiple attacks; routing attacks, and side-channel attacks at the infrastructure layer and Replay attacks.

Complex Approach in Cryptanalysis of Internet of Things (IoT)

61

Talking about the blockchain we need to notice that the network represents a unique chain representing a consortium of companies with interconnected IoT devices, or a smart city. The blockchain networks are growing and expanding, getting increasingly needed to interact to each other. This kind of interaction requires new interoperability and security standards and protocols, and toolkits to provide high level security for the discrete blockchain bridges (Fig. 6). Blockchain architecture is quite different from a client/server architecture (Fig. 7) (Fig. 8).

Fig. 6. Multi-provider blockchain networks with open standards and global interoperability [4].

Fig. 7. Centralized client/server architecture/Decentralized blockchain architecture [5].

62

L. Mirtskhulava et al.

Fig. 8. Cisco Network enabled blockchain

4 Lattice-Based Reduction Algorithms In our cryptanalysis we’re using lattices capable to break systems. A lattice L is a set of basis vectors {b1 , . . ., bm } taking the integer linear bi combinations.

Fig. 9. A lattice with two base vectors.

Complex Approach in Cryptanalysis of Internet of Things (IoT)

63

B is called a basis matrix. We consider lattice L formed by the two vectors. For example, if we have the two vectors like     2 1 b1 = b2 = 2 1 So, forming two-dimensional lattice we have result presented on Fig. 9.

5 The NTRU Cryptosystem In NTRU cryptosystem we use three integer parameters N, p, q and the set of four Lf , Lg , L∅ , Lm polynomials of with integer coefficients and N-1 degree. We note that p and q should not be prime and we assume that gcd(p; q) = 1, where q consider to be larger than p. We have the ring R = Z[X]/(Xˆ(N−1)) and the element F∈R could be written as a polynomial or a vector, F=

N −1 

  Fi xi = F0 , F1 , . . . , FN −1

i=0

We denote multiplication in R using the symbol * given explicitly as a cyclic convolution product, F ∗G =H where Hk =

k  i=0

N−1 

Fi Gk−i +

Fi GN+k−i =

i=k+1



F i Gi

i+J≡k(modN)

By using a multiplication modulo q, we can reduce the coefficients modulo q. First step: key generation. For generating the NTRU key, Bob chooses 2 polynomials fg∈Lg randomly. The polynomial f should satisfy the following requirement that means it has inverse modulo q and modulo p. For choosing the right parameters, what is true in most choices of f. To compute these inverses is quite easy to use the modification of the Euclidean algorithm. Let’s denote the inverses by Fq and Fp, as a result, we have Fq ∗ f ≡ 1(modq) and Fp ∗ f ≡ 1(modp) Then Bob computes the quantity h ≡ Fq ∗ g(modq) The public key of Bob represents the h polynomial. The private key of Bob represents the f polynomial, and in practice he would want to store Fp.

64

L. Mirtskhulava et al.

Second step: encryption: We suppose that Alice wants to send a message to Bob. She first begins to select a message m from the set of plaintexts Lm. Next she chooses a polynomial L randomly and uses Bob’s public key h to compute that means the message which is encrypted and transmitted by Alice to Bob. Third step: decryption: We suppose that e message was received by Bob from Alice and after that it could be decrypted by using his private key f. The best way to do that, Bob has to precompute the Fp polynomial. For decrypting e, Bob’s first step is to compute. Next step of Bob is to choose the coefficients from interval [−q/2, q/2]. Next step is to treat a as the polynomial with integer coefficients then Bob will recover the message and compute the key.

6 Conclusions In the given paper, we proposed Cisco blockchain framework as the best solution for securing IoT. Cisco hardware-independent blockchain framework specifies a set of standards addressing infrastructure-level security risks based on reference architectures deploying on the premises: in an demilitarized zone (DMZ) of an enterprises, and in the cloud or a hybrid deployment. The proposed platform supports hardware security modules (HSMs) and WAN optimization tools. We analysed NTRU-based cryptosystem: encryption and authentication schemes for securing IoT wireless connection. This approach is an alternative solution to support data integrity, confidentiality, authentication issues. NTRU was offered by evaluation of other asymmetric algorithms such as RSA, ECC. Main advantage of the proposed cryptosystems is their encryption, decryption and key-generation speeds as they are faster than the others.

References 1. Recommendation ITU-T Y.2060: Overview of the Internet of Things, June 2012 2. Mirtskhulava, L., Gulua, N., Meshveliani, N.: Ntru cryptosystem analysis for securing IoT. GESJ: Comput. Sci. Telecommun. 1(56) ( 2019). ISSN 1512-1232 3. Rayes, A., Salam, S.: Internet of Things from Hype to Reality: The Road to Digitization, 2nd edn. Springer, Cham (2019) 4. Blockchain by Cisco. Build trust-based business networks for digital transformation (2016) 5. Deploying Enterprise: BlockchainsRam Jagadeesan CTO Blockchain. Cisco (2019) 6. https://www.uk.sogeti.com/content-hub/blog/iot-security-using-blockchain/ 7. https://www.forbes.com/sites/kateoflahertyuk/2019/04/11/flaws-in-wpa3-wi-fi-standardallow-attackers-to-crack-passwords-and-view-traffic/#6df21b617050 8. Rayes, A., Salam, S.: Internet of Things from Hype to Reality: The Road to Digitization, 2nd edn. Springer, Cham (2019) 9. Smart, N.P.: Cryptography Made Simple. Springer, Cham (2016) 10. Sklavos, N., Zhang, X.: Wireless Security and Cryptography, Specifications and Implementations. Taylor & Francis Group, LLC. (2006). p. 404 11. Silverman, J.H., Hoffstein, J., Pipher, J.: NTRU: a ring based public key cryptosystem. In: Algorithmic Number Theory, pp. 267–288. Springer, Heidelberg (1998) 12. Mekhazniaa, T., Zidania, A.: Wi-Fi security analysis. Procedia Comput. Sci. 73, 172–178 (2015)

Complex Approach in Cryptanalysis of Internet of Things (IoT)

65

13. Menezes, A.J., van Oorschot, P.C., Vanstone, S.A.: Handbook of Applied Cryptography. Fifth Printing. CRC Press, August 2001. 816 pp., ISBN 0-8493-8523-7 14. Sakib, N., Ahmed, S., Rahman, S., Mahmud, I., Belali, Md.H.: WPA 2 (Wi-Fi Protected Access 2) Security Enhancement: Analysis & Improvement. Global Journal of Computer Science and Technology vol. 12 Issue 6 Version 1.0 March 2012. Online ISSN: 0975-4172 & Print ISSN: 0975-4350. USA 15. Mwangi, J., Cheruiyot, W., Kimwel, M.: Security analysis of WPA2. Control Theory Inform. 5(5) (2015). ISSN 2224-5774 (Paper) ISSN 2225-0492 (Online) 16. Rivest, R.: The RC4 encryption algorithm. RSA Data Security (1992) 17. Microsoft Technet Library, How 802.11 Wireless Works, Technical Reference. http://technet. microsoft.com/enus/library/cc757419(WS.10).aspx 18. Arana, P.: INFS 612 – Fall 2006 “Benefits and Vulnerabilities of Wi-Fi Protected Access 2 (WPA2)” 19. Moen, V., Raddum, H., Hole, K.J.: Weaknesses in the temporal key hash of WPA. Mob. Comput. Commun. Rev. 8, 76–83 (2001) 20. Vanhoef, M., Piessens, F.: Key reinstallation attacks: forcing nonce reuse in WPA2. https:// papers.mathyvanhoef.com/ccs2017.pdf 21. http://searchsecurity.techtarget.com/feature/Control-wireless-vulnerabilities-before-theycontrol-you 22. Beaver, K., Davis, P.T., Akin, D.K.: Hacking Wireless Networks for Dummies 2005. ISBN 978-0-7645-9730-5 23. https://en.wikipedia.org/wiki/IEEE_802.11i-2004 24. https://arstechnica.com/information-technology/2017/10/severe-flaw-in-wpa2-protocol-lea ves-wi-fi-traffic-open-to-eavesdropping/ 25. Silverman, J.H., Hoffstein, J., Piper, J.: NSS: an NTRU lattice-based signature scheme. In: EUROCRYPT, Lecture Notes in Computer Science. Springer, Heidelberg (2001). ISBN 9783-540-42070-5 26. Singh, S., Padhye, S.: Generalisations of NTRU cryptosystem. Security Commun. Networks 9, 315–6334 (2016) 27. Coglianese, M., Goi, B.M.: MaTRU a new NTRU-based cryptosystem. In: INDOCRYPT 2005, LNCS (2005) 28. Akleylek, S., Kaya, N.: New quantum secure key exchange protocols based on MaTRU. IEEE Xplore, digital library. In: 2018 6th International Symposium on Digital Forensic and Security (ISDFS), 22–25 March 2018

Biometric Cryptosystems: Overview, State-of-the-Art and Perspective Directions Maria Lutsenko1(B) , Alexandr Kuznetsov1(B) , Anastasiia Kiian1(B) Oleksii Smirnov2(B) , and Tetiana Kuznetsova1(B)

,

1 V. N. Karazin, Kharkiv National University, Svobody Square 4, Kharkiv 61022, Ukraine

[email protected], [email protected], [email protected], [email protected] 2 Central Ukrainian National Technical University, Avenue University, 8, Kropivnitskiy 25006, Ukraine [email protected]

Abstract. Modern cryptographic systems are constantly evolving and improving. This is due to the development of new computing systems and advanced cryptographic analysis methods, as well as increasing requirements for speed, security and reliability of used tools. In particular, it was announced an advent of universal quantum computers, which will be able to provide cryptanalysis with advanced calculation methods based on fundamentally new physical principles. The possible use of such devices encourages the development, research and standardization of algorithms for post-quantum cryptographic information protection. Another factor for the development of advanced cryptographic systems is the biometric technologies popularization. In this work, a critical review and analysis of the current application of biometric technologies in cryptographic systems, is conducted. In particular, biometric cryptographic systems, which are designed to generate secure pseudorandom sequences that can be used as cryptographic keys, passwords etc., are investigated. A comparative analysis of various biometric cryptosystems with the determination of their advantages and disadvantages is carried out. The perspective directions for further research are substantiated. Also this work presents a new key generation scheme which uses fuzzy extractors from the biometric data of iris. The proposed method is based on the code-based public key cryptosystems which are considered to be resistant to quantum cryptanalysis. A software implementation of this method with experimental studies of the key generation algorithm and recommendations on the practical application are proposed. Keywords: Biometric cryptosystems · Fuzzy extractors · Biometric data · Code-based public key cryptosystems

1 Introduction At the moment, the question of combining classical cryptography with biometric technology is relevant. Mathematical models and information security methods based on © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 66–84, 2021. https://doi.org/10.1007/978-3-030-58359-0_5

Biometric Cryptosystems

67

the use of biometric data become one of the main elements in providing highly reliable identification and verification systems [1–4]. Biometric authentication enables the implementation of key data protection mechanism using unique biometric features. In addition, biometric data can be a complete substitute for cryptographic keys by generating stable and robust pseudorandom sequences. Depending on the purpose of biometrics application in cryptography there are several types [4–7] of biometric cryptographic systems (as shown on Fig. 1):

Biometric Key Cryptosystems Key Release Cryptosystems Biometric-based authentication

Key Binding Cryptosystems Biometric encryption

Key Generation Cryptosystems Fuzzy extractor

Fuzzy commitment scheme

Private template scheme

Fuzzy vault

Quantization scheme

Shielding function

Fig. 1. Types and subspecies of biometric cryptographic systems

• Key Release Cryptosystems; • Key Binding Cryptosystems; • Key Generation Cryptosystems. In sections below will be conducted analysis and comparative studies of biometric cryptographic systems, justification of perspective directions of their development and possible application.

2 Biometric Key Release Cryptosystems In key release mode, biometric authentication is performed regardless of the key release mechanism, the biometric reference and cryptographic key are stored separately from each other, the key itself is released after successful biometric authentication. In this case, the cryptographic key’s password is the result of comparing the received biometric image with the template [4–7]. A schematic representation of such system is shown in Fig. 2.

68

M. Lutsenko et al.

Formation of a cryptographic key

Combining the key with biometrics

User registration

Biometric data Biometric template Key Secure storage

Biometric data Comparison of biometrics with a template

Key release

Key release

Fig. 2. Schematic representation of biometric cryptographic system with key release

3 Biometric Key Binding Cryptosystems In cryptographic systems of this type, the key and the biometric reference are cryptographically interconnected [4–7]. The key by a particular algorithm is associated with the biometric reference of the user and stored as it is in the database, so it is only possible to disclose the key to the owner of the biometric parameters. Such systems do, however, not require the use of helper data to unmask noisy biometric data. A schematic example is shown in Fig. 3. User registration

Formation of a cryptographic key

Key binding with biometrics Biometric data

Helper data

Secure template

Public storage

Secure storage Biometric data

Key release

Key Key extraction

Fig. 3. Schematic representation of a biometric cryptographic system with key binding

This method involves hiding the cryptographic key in the biometric registration template itself with the help of a reliable (secret) bit replacement algorithm. However, if an attacker was able to identify the bit locations that specifies the key, then the attacker can recover the embedded key from any other user’s template.

Biometric Cryptosystems

69

3.1 Biometric Encryption Biometric encryption is a process that securely associates a cryptographic key with a biometric data, because neither the key nor the biometric data can be retrieved from the stored template [4–7]. The cryptographic key is randomly generated at registration so that no one, including the user, knows it. The key itself is completely independent of biometrics and therefore can always be changed or updated. After receiving the biometric sample, a secure template is created, called a “private template”. Essentially, a cryptographic key is encrypted using a biometric data. A schematic example is shown in Fig. 4.

Cryptographic key

Biometric data Template

User registration

Encryption Lookup table

Random sequence Public storage

Template

Biometric data

Identification code Hash

Decryption

Private template

Secure storage

Cryptographic key Key extraction

Fig. 4. Schematic representation of biometric cryptographic system with key binding: biometric encryption scheme

3.2 Fuzzy Commitment Scheme Fuzzy Commitment Scheme is a cryptographic algorithm that provides biometric data storage using cryptography and error correction coding methods. The algorithm associates secret information with data to hide data and not allow the data owner to reveal it in more than one way [6, 8, 9]. This scheme, which applies to biometric templates, treats the template itself without any modification as a corrupted sequence to be decoded. In such systems, a so-called witness (or encryption key) is used to form the template and retrieve the data. It is also believed that for the proper functioning of the algorithm, witness may have similar but not identical metrics. Schematic representation is shown in Fig. 5.

70

M. Lutsenko et al. User registration

Error correction coding

Biometric data

Witness

Binding

Codeword

The hash result

Difference vector Public storage

Secure storage

Witness

Restoring Codeword

Biometric data

Key extraction

Fig. 5. Schematic representation of a biometric cryptographic system with key binding: fuzzy commitment scheme

3.3 Fuzzy Vault In their work, Ari Juels and Madhu Sudan, in 2002 [8], described a new design, which they called a fuzzy vault. The main idea is summarized as follows. If a person tries to unlock some vault using own biometric set, he will succeed if the sets of it and the reference sets intersect significantly. For this purpose, biometric data are processed using cryptography and error-coding methods. Thus, fuzzy vault can be considered as an error checking scheme in which the keys consist of data sets. Schematic representation of the biometric cryptographic system is shown in Fig. 6.

Error correction coding

Biometric data

Characteristics set

Characteristic polynom

User registration Secret

Vault Chaff points

Template

Public storage Secure storage

Characteristics set Biometric data

Characteristic polynom Secret

Key extraction

Fig. 6. Schematic representation of biometric cryptographic system with key binding: fuzzy vault scheme

Biometric Cryptosystems

71

3.4 Shielding Function Shielding function or helper data scheme has been developed to ensure the safety of stored biometric data [4–6]. This approach allows the system to verify the identity of the user without any knowledge of the biometric data. Delta-contracting and epsilondetecting functions provide the basis for this scheme. The first function binds the secret to the biometric data, and the epsilon-detection feature ensures that the private template reveals only a small amount of secret and biometrics data. A schematic representation is shown in Fig. 7. User registration

Error correction coding

Biometric data Characteristic encoding

Cryptographic key Binding Hash Private templete

Characteristic encoding

Secure storage

Releasing Error correction coding

Biometric data

Cryptographic key

Key extraction

Fig. 7. Schematic representation of biometric cryptographic system with key-linking: shielding function scheme

3.5 Comparative Analysis of Biometric Cryptosystems with Key Binding A comparison of the advantages and disadvantages of biometric cryptographic systems with key binding is given in Table 1. Table 1. Comparative analysis of biometric cryptographic systems with key binding Scheme

Advantages

Disadvantages

Biometric encryption

Applies a classic cryptographic algorithm to create a biometric template. An attacker is unable to decrypt private templates without knowing the algorithm and the cryptographic key

It is possible to use a reconstructed secret to retrieve biometrics from a template. Not resistant to blended substitution attacks, masquerade attacks and nearest impostor attacks, etc. (continued)

72

M. Lutsenko et al. Table 1. (continued)

Scheme

Advantages

Disadvantages

Fuzzy commitment scheme “Commitment” derived from biometric data and a private key protect the biometric template. Also, the private key is protected because only its hash value is stored

Vulnerable to all known encryption attacks (depends on the encoding algorithm selected). Not resistant to climbing attacks, brute force, decoding, and crossmatching attacks

Fuzzy vault

The vault cannot be decoded without biometric data having almost identical characteristics to the primary data

Susceptible to brute force attacks and collisions. Vulnerable to intrusion, liaison, combined and injection attacks

Shielding function

Helper data and hash protect biometric data and secrets from reproduction. Biometric data cannot be restored from a private template without the private key

Short length of keys. Ability to use re-engineered secret to retrieve biometric from compromised helper data. Not resistant to brute force, crossmatching attacks, attact via record multiple

4 Biometric Cryptosystems with Key Generation In such a biometric cryptosystem, the key is generated directly from the user’s biometric data and is not stored in the database. The ability to not store a key derived from biometric data is an undeniable advantage. A schematic representation of such a system is shown in Fig. 8. Key generation

Biometric data

Key generation

Helper data

Key

Public storage

Biometric data

Key Recreating

Recovery of key

Fig. 8. Schematic representation of a biometric cryptographic system with key generation

Such systems are more secure [3–9], but difficult to apply because of even small variability in biometric, since it is necessary to generate the same key over and over again from approximately similar data. The disadvantage of such systems is the inability (or

Biometric Cryptosystems

73

limited capacity) to generate a new key. Therefore, if a cryptographic key is ever compromised, it will not be possible to use this particular biometric image and a specific key generation algorithm. On a system that needs a cryptographic key update periodically, this is unacceptable. 4.1 Fuzzy Extractor The most common technology underlying biometric cryptographic systems with key generation are fuzzy extractors [6]. The basic logic behind using fuzzy extractors is similar to fuzzy vault. This method allows to uniquely recover the private key from inaccurate (noisy) biometric data. The method of fuzzy extractors involves the formation of a randomly distributed sequence from the original data and further correct its recovery from any data quite similar to the original. Initially, a bit sequence is generated, which is encoded by a error correcting code. Hamming, Adamar, Bose–Chaudhuri–Hocquenghem, Reed - Solomon codes can be used [1–3]. The generated bit sequence can be designed to identify, authenticate or generate cryptographic encryption keys. The described method is schematically depicted in Fig. 9. Key generation Private key

Biometric data Error correction coding

Combine

Error correction coding

Extract

Private key

Template

Public storage

Biometric data

Recovery of key

Fig. 9. Schematic representation of the fuzzy extractor scheme

The two main approaches used to generate biometric keys are the private template scheme and the quantization scheme. 4.2 Private Template Scheme Private template scheme uses helper data, a sequence of validation bits, to correct errors [10]. The key itself is formed directly from the biometric image or from the hash of that biometric image. Here is presented an example of a private template schema in which a key is formed from a hash. A schematic representation of such an algorithm is shown in Fig. 10.

74

M. Lutsenko et al.

Formation of characteristic vector

Majority decoding

Key generation

Biometric data Error correction coding

Helper data

Key

Public storage

Biometric data

Majority decoding

Hash

Formation of characteristic vector Recovery of key

Fig. 10. Schematic representation of the cryptosystem with key generation: private template scheme

4.3 Quantization Scheme Quantization schemes generate biometric keys using helper data and binarized (or quantized) biometric characteristics. A unique feature of quantization schemes is the ability to retrieve the same keys from quite a variety of biometric images, even if obtained using different scanners. Schematic functioning is presented in Fig. 11.

Characteristic Extractor

Key generation Defining intervals

Biometric data Intervals

Reflection of intervals

Helper data Key Interval encoding Public storage

Biometric data

Characteristic Extractor

Hash

Reflection of intervals Recovery of key

Fig. 11. Schematic representation of cryptosystem with key generation: quantization scheme

For the operation of the quantization scheme, the characteristic vectors of several biometric samples are required to calculate the corresponding intervals for each characteristic element [4–10]. These intervals are stored as helper data. In order to provide cancelable keys, most schemes provide parameterized interval encoding.

Biometric Cryptosystems

75

4.4 Comparative Analysis of Biometrıc Cryptosystems with Key Generation A comparison of the advantages and disadvantages of cryptographic methods is given in Table 2. Table 2. Comparative analysis of biometric cryptographic systems with key generation Scheme

Advantages

Disadvantages

Private template scheme

Generates keys directly from Keys are not cancelable in case biometric data. Provides template of compromise protection. The use of strong hash algorithms to generate random keys from biometric data provides cryptographic resistance to brute force attacks

Quantization scheme

It is possible to recover the same keys from multiple copies of biometric data, obtained even from different scanners. Provides security and privacy by not storing biometric data

Vulnerable to attack via multiple records

5 Comparative Analysis of Biometrıc Cryptosystems Let analyze the advantages and disadvantages of the above biometric cryptographic systems. The results are shown in Table 3. Table 3. Comparative analysis of biometric cryptographic systems Scheme

Advantages

Disadvantages

Key Release Crypto systems

Ease of practical implementation More secure password replacement for keys. The key is generated by a reliable cryptographic generator

The template must be stored in a database, which means that it can be stolen. The cryptographic key must be stored as part of the template. Does not guarantee a high level of data security (continued)

76

M. Lutsenko et al. Table 3. (continued)

Scheme

Advantages

Disadvantages

Key Binding Crypto systems

The key is stored in a encoded form. The key is generated by a reliable cryptographic generator. Helper data does not contain any biometric images and key, so it can be stored in the open storage

The template must be stored in a database, which means that it can be stolen. The cryptographic key must not be stored as part of the template

Key Generation Crypto systems

Key generation directly from biometric templates. Helper data does not contain any biometric image and key, so it can be stored in the open storage. The generated key does not contain any owner’s biometric data

The biometric characteristics do not provide sufficient information to obtain a reliable, renewable key without using any up-data. Difficulty in creating a new key when compromising a previous one (human limitation as a biometric image carrier). Vulnerability to attacks by brute force and masquerade attacks, false identity attacks

6 Development of the Scheme of Biometric Data Extracting by Fuzzy Extractors from Iris 6.1 Fuzzy Extractors By definition, a (M , m, l, t, e)-fuzzy extractor is a pair of randomized procedures, generation Gen and reproduction Rep. The fuzzy extractor is effective, if Gen and Rep performed in less than polynomial time. In other words, fuzzy extractors allow to extract some randomness R from w, and then successfully restore R from any w close to w. The restore uses the helping data P that is generated during the initial extraction; however, P do not necessarily have to be kept secret, since R are close to a truly random sequence, even if P known. Strong extractors can indeed be considered as “clear” analogues of fuzzy extractors that are relevant if one assumes that t = 0, P = X and M = {0, 1}n . 6.2 Biometric Data Extracting from the Iris The iris is recognized as one of the most reliable, unique and non-invasive biometric methods for obtaining biometric person’s data [10–21]. It is believed that this person’s metric does not depend on human genetics, and therefore guarantees a unique identification of a person, that the probability of coincidence of two iris among different people is equal 2−78 . In this work, all images were taken from the CASIA Iris Image Database. Segmentation. Segmentation is the process of finding the most useful part of an iris‘s image for further processing. This is done by locating the pupil and the border of the

Biometric Cryptosystems

77

iris and eyelids. In the absence of proper segmentation, the subsequent stages of iris recognition will be incorrect due to the large number of third-party data (noise), resulting in erroneous data being generated as a template. For segmentation of the iris [10–21], it is possible to localize it with a simple combination of Gaussian filtering, the discovery of the iris edge by the Canny edge detection operator and the Hough transform transformation. Canny edge detection operator. The algorithm for detecting the boundaries of the iris with the help of the Canny operator can be divided stages as shown on Fig. 12.

Apply the Gauss filter to smooth the image to remove the noise and remove undesirable details of the image

Find gradients of the intensity of the image. To do this, can use Sobel gradient operator.

Apply the non-maximum suppression to get rid of false data on the detected edge

Apply a double threshold to determine potential boundaries. The edge obtained in the previous step needs to be clarified.

Edge Tracking by Hysteresis. Isolate the edge of the iris, suppressing all other edges that are weak and not related to the strong edges of the iris.

Fig. 12. Schematic representation of applying Canny edge detection operator

Hough transform. Used to determine the radius and center of the pupil and iris. As shown in Fig. 13, the localized iris is normalized to a rectangular block with a fixed size corresponding to the width of the block, and the angular displacement  corresponds to the length of the block. An example of normalizing the iris is shown in Fig. 14.

78

M. Lutsenko et al. Θ 0

r

Iris biometric data

r

1

Θ Eye: iris and pupil

Fig. 13. Schematic representation of the normalization of the image of the iris

Input image

Segmented image

Normalized image

Fig. 14. Formation of a normalized image of the iris

The most common way of normalizing is to construct a Daugman’s rubbersheet model. Formally, normalization is a linear model, which is formed from the corresponding pixel of the iris, regardless of its size and the state of pupil expansion, a pair of polar coordinates (r, ), where r is located on the unit interval r ∈ [0, 1] and  is an angle in the range  ∈ [0, 2π ]. Data Formation. Data formation is a process of extracting information from an iris image. During the removal of the iris data from a normalized image, a control area is allocated. Gabor filters [7, 10] are used for each point of the selected area in order to extract phase information. The undoubted advantage of the phase information is that it, unlike the amplitude information, does not depend on the contrast of the image and the illumination. It should be noted that these data can’t be used to reconstruct person’s iris. The Gabor filter is a linear filter, which is a convolution of Fourier transforms of the harmonic function and Gaussian function. Let denote f the frequency of a sinusoidal plane wave, α and β - the space constants of the Gaussian wavelets on the axis x and y ,  - the orientation of the Gabor filter. In this case, the 2D Gabor filter is defined as: G(x, y, ) =

1 e 2π αβ

 2   2    − x 2 − y 2 2α

2β

  cos 2π fx

(1)

where x = x sin  + y cos , y = −x cos  + y sin . Therefore, the procedures for forming the biometric data can be presented as shown in Fig. 15. After the stage of the formation of biometric data begins the stage of cryptographic key generation.

Biometric Cryptosystems

79

Fig. 15. Schematic representation of the proposed biometric cryptographic system with key generation

6.3 Cryptographic Key Generation from the Iris Biometric Data Using Fuzzy Extractors The key generation stage consists of two components: biometric data error-correcting encoding and hashing the resulting sequence. Error-correcting encoding is required directly, as the implementation of the method of fuzzy extractors. Hashing procedure is optional. It is necessary in case of compromising the key to not compromise the biometric data or to make it impossible to use a system by a certain person. Error-correcting encoding done by Reed-Solomon codes. Reed-Solomon’s codes are cyclic codes that correct errors. At present, the Reed-Solomon codes have a very wide scope of application due to their ability to find and correct multiple mistakes. Data hashing with the Kupina algorithm (DSTU 7564:2014). Kupina is an iterative cryptographic function of hashing which was adopted as the national standard of Ukraine DSTU 7564:2014 [22, 23]. Formation of a key is carried out by the method of fuzzy extractors according to the scheme of the private template. The algorithm for generating the key is schematically presented in Fig. 15. To improve the performance of the algorithm, the helping data bits of the Reed-Solomon codes for correcting errors will be used. The Key Generation Algorithm: • For the biometric data T of the length of the M bit, the coding of Reed-Solomon codes is used. As a result, an error correction vector Vec(C) is generated Vec(C) =   (C1 , C2 , . . . , Cn ), for n-bit code defined as Vec(ci ) = ci,1 , ci,2 , . . . , ci,n . • Then, the checksum vector is combined with the biometric data Vec(C) || Vec(T ). • In general, the sequence after encoding has the form Vec(C, T ) = (C1 , C2 , . . . , Cn || T1 , T2 , . . . , TM ).

80

M. Lutsenko et al.

• To form the key, the hashing algorithm Kupina-512 for sequence Vec(C)||Vec(T ) is executed. Key = Hash[Vec(C) || Vec(T )]

(2)

• As a result, a sequence of 512 bits is formed which is suitable for use as a cryptographic key. Key Reproduction Algorithm: • To the biometric data T  the length of the M bit is accompanied by  an error   C, T  = correction vector that was formed on key generation stage, so Vec    . C1 , C2 , . . . , Cn || T1 , T2 , . . . , TM  • By the received sequence Vec C, T  , decoding is used by the Reed-Solomon  algorithm. After successful decoding the vector    Vec T is formed. • For key generation hashing Vec(C) || Vec T are performed    Key = Hash Vec(C) || Vec T  (3)

7 Software Implementation and Experimental Studies 7.1 Input and Output Data At the stage of extracting biometric data input data will be an image (two-dimensional data array) of an eye of a certain size. In this case, as were used eye images contained in the CASIA library, the size of the eye image will be 320 × 280 pixels. Given that receiving the full image of the iris is a rather difficult task, since in the normal state of the human eye is incompletely open, the iris can partly overlap the upper and lower eyelids, the eyelashes. Approximately 3/4 of the iris can be captured without inconvenience to the user. Therefore, it is only appropriate to consider 6,144 elements of the vector obtained after applying the Gabor filter. Thus, the initial data of the phase of data extraction   is a vector VecGABOR (BD) = (BD1 , BD2 , . . . , BDM ) with elements BDi ∈ GF 28 , i = 1, 2, . . . , M , M = 6144. The input data for the key generation stage is a bit array of L = M × k = 6144 × 8 = 49152 bit length that was generated from the vector VecGABOR (BD). The output data is a bit sequence of 512 bits in length - a cryptographic key. 7.2 An Example of Using the Software Implementation The stage of extracting data from the eye image is shown in Fig. 16.

Biometric Cryptosystems

81

Iris image

Canny operator: determining the external edges of the iris Hough transformation: the definition of the inner edges of the iris

Gauss filter: select areas of the image that do not contain the iris

A certain threshold mechanism that rejects the noisy images of the iris

Construction of the Daugman’s rubbersheet model: polar transformation

2D Gabor filter: forming an array of biometric data

Biometric data

Fig. 16. Results of application of the proposed processing scheme for extracting biometric data from the iris at the data extraction stage

7.3 Experimental Research of Key Generation Algorithm For the main transformations in both stages of implementation, the following performance indicators were obtained, as shown in Table 4. To evaluate the performance, a calculation of the execution time of the main operation was performed during the processing of each of the 756 images from the CASIA database. Performance analysis was performed on a computing platform running Windows 10 x64, Intel Core i7, 4.7 GHz. We evaluate the proposed algorithm for parameters such as the probability of False Accept Rate (FAR) and the probability that the system does not recognize the authenticity of the biometric data registered in its user - False Reject Rate (FRR).

82

M. Lutsenko et al.

Table 4. Evaluation of performance of the proposed software implementation of key generation algorithm Operation

Performance, ms

Localization of the iris Canny edge detection operator 43,64 Hough transform

826

Isolation of the iris

2,9

Normalization of the iris

6,58

Retrieving data

113,09

Reed-Solomon codes Encoding

1,56

Decoding

2,14

10000 8000 6000 4000 2000 0 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 11% 12% 13% 14% 15% 16% 17% 18% 19% 20% 21% 22% 23% 24% 25%

Number of experiments

Consequently, it was found that the probability of erroneous identification for the proposed algorithm is quite low FAR = 0.14%. It is established that the level of FRR for the proposed algorithm is 19.5%. This is a fairly high rate of efficiency of the algorithm, but the urgent issue is to increase the level of reliability of the implementation. On Fig. 17 shows the dependence of the FRR level on the correcting code capability for the proposed implementation.

Remedial ability, in percentages

Fig. 17. Dependence of the probability of successful decoding from the remedial ability of the codes of Reed-Solomon

To perform this analysis, a random binary sequence was formed, then it was coded by the Reed-Solomon codes, the correction vector memorized as a helping data, and then the input sequence was changed in accordance with the capabilities of the correcting code (codes are considered which can correct from 1% to 25% of errors). Then the sequence was decoded. This procedure was repeated 10,000 times. The histogram shows the number of successful decoded corrupted sequences. Based on the results obtained, we can conclude that the search for an optimal algorithm for jamming coding is also a promising direction for further research.

Biometric Cryptosystems

83

8 Conclusions Methods for extracting biometric data have long been known. However, it is at the stage of creating post-quantum cryptography that biometric cryptographic systems can find their widespread use and become a more reliable substitute for not only passwords, but also the crypyographic keys. In this paper, three types of biometric cryptosystems are considered. However, biometric cryptosystems with key generation are the most promising areas of research in terms of cryptographic stability. In this paper, an algorithm for extracting biometric data from the iris based on the scheme of fuzzy extractors is proposed. The proposed algorithm consists of two steps: data extraction and cryptographic key generation. The method are based on information theory and error corection coding. The use of the fuzzy extractor method can compensate for errors arising from the technical inability to obtain the same values of biometric characteristics. The fuzzy extractor scheme allows you to get a key that meets all the criteria for cryptographic key security. Also in the work, an estimation of performance indicators of the developed algorithm was conducted. Performing a complete procedure for data extraction and key generation takes less than 1 s. The FAR and FRR are, respectively, 0.14% and 19.5%.

References 1. Jain, A.K., Bolle, R., Pankanti, S.: Biometrics: Personal Identification in Networked Society, vol. 1 (1999). 434 p. 2. Kuznetsov, A., Kiyan, A., Uvarova, A., Serhiienko, R., Smirnov, V.: New code based fuzzy extractor for biometric cryptography. In: 2018 International Scientific-Practical Conference Problems of Infocommunications. Science and Technology (PIC S&T), Kharkiv, Ukraine, pp. 119–124 (2018). https://doi.org/10.1109/INFOCOMMST.2018.8632040 3. Rassomakhin, S., Kuznetsov, A., Shlokin, V., Belozertsev, I., Serhiienko, R.: Mathematical model for the probabilistic minutia distribution in biometric fingerprint images. In: 2018 IEEE Second International Conference on Data Stream Mining & Processing (DSMP), Lviv, Ukraine, pp. 514–518 (2018). https://doi.org/10.1109/dsmp.2018.8478496 4. Uludag, U., Pankanti, S., Prabhakar, S., Jain, A.: Biometric cryptosystems: issues and challenges. Proc. IEEE 92(6), 948–960 (2004) 5. Jain, A.K., Ross, A.: An introduction to biometric recognition. IEEE Trans. Circuits Syst. Video Technol. 14(1), 4–20 (2004) 6. Dodis, Y., Ostrovsky, R., Reyzin, L., Smith, A.: Fuzzy extractors: how to generate strong keys from biometrics and other noisy data. SIAM J. Comput. 38(1), 97–139 (2008) 7. Wu, L., Liu, X., Yuan, S., Xiao, P.: A novel key generation cryptosystem based on face features. In: 2010 IEEE 10th International Conference on Signal Processing (ICSP), pp. 1675–1678 (2010) 8. Juels, A., Sudan, M.: A fuzzy vault scheme. Des. Codes Cryptogr. 38(2), 237–257 (2006) 9. Boyen, X.: Reusable cryptographic fuzzy extractors. In: 11th ACM Conference on Computer and Communications Security, USA, pp. 82–91 (2004) 10. Rathgeb, C., Uhl, A., Wild, P.: Iris-biometrics: from segmentation to template security. Advances in Information Security, Springer (2013) 11. Daugman, J.: How Iris recognition works. IEEE Trans. Circuits Syst. Video Technol. 14(1), 21–30 (2004)

84

M. Lutsenko et al.

12. Hao, F., Anderson, R., Daugman, J.: Combining crypto with biometrics effectively. IEEE Trans. Comput. 55, 1081–1088 (2006) 13. Wildes, R.P.: Iris recognition: an emerging biometric technology. Proc. IEEE 85, 1348–1363 (1997) 14. Boles, W.W., Boashash, B.: A human identification technique using images of the iris and wavelet transform. IEEE Trans. Sig. Process. 46(4), 1185–1188 (1998) 15. Lim, S.L., Lee, K.L., Byeon, O.B., Kim, T.K.: Efficient iris recognition through improvement of feature vector and classifier. ETRI J. 23(2), 61–70 (2001) 16. Davida, G., Frankel, Y., Matt, B.: On the relation of error correction and cryptography to an off-line biometric identification scheme. In: Proceedings of Workshop on Coding and Cryptography, Paris, France, pp. 129–138 (1999) 17. Bae, K., Noh, S., Kim, J.: Iris feature extraction using independent component analysis. In: 4th International Conference on Audio-and Video-based Biometric Person Authentication, Guildford, UK, pp. 838–844 (2003) 18. Tisse, C., Martin, L., Torres, L., Robert, M.: Person identification technique using human iris recognition. In: Proceedings of the Vision Interface, pp. 294–299 (2002) 19. Ma, L., Tan, T., Wang, Y., Zhang, D.: Efficient Iris recognition by characterizing key local variations. IEEE. Image Process. 13, 739–750 (2004) 20. Ogiela, M.R., Ogiela, L.: Image based crypto-biometric key generation. In: 2011 Third International Conference on Intelligent Networking and Collaborative Systems, Fukuoka, Japan, pp. 673–678 (2011) 21. Chawla, S., Oberoi, A.: A robust algorithm for iris segmentation and normalization using Hough transform. Global J. Bus. Manag. Inf. Technol. 1(2), 69–76 (2011) 22. A New Standard of Ukraine: The Kupyna Hash Function. https://eprint.iacr.org/2015/885. pdf. Accessed 12 Sept 2015 23. Dobraunig, C., Eichlseder, M., Mendel, F.: Analysis of the kupyna-256 hash function. Lecture Notes in Computer Science, pp. 575–590 (2016)

Biometric Authentication Using Convolutional Neural Networks Alexandr Kuznetsov1(B) , Inna Oleshko2(B) , Kyrylo Chernov1(B) Mykhaylo Bagmut1(B) , and Tetiana Smirnova3(B)

,

1 V. N. Karazin Kharkiv National University, Svobody Square 4, Kharkiv 61022, Ukraine

[email protected], [email protected], [email protected] 2 Kharkiv National University of Radio Electronics, Nauky Avenue 14, Kharkiv 61166, Ukraine [email protected] 3 Central Ukrainian National Technical University, Avenue University, 8, Kropivnitskiy 25006, Ukraine [email protected]

Abstract. Today biometric identity authentication technologies are widespread. These systems are implemented not only in enterprises, controlled-access facilities, but also on smartphones of ordinary users and in online applications. The problem of choosing one of the authentication methods remains urgent. This paper provides a comparative analysis of existing systems and concludes that one of the most common and persistent methods is facial authentication system. The most powerful types of attacks on the biometric system are attacks on the database of biometric templates and attacks on sensors for obtaining biometric characteristics. Attacks on biometric sensors or spoofing attack is aimed at impersonating another person through fake biometric data. The paper deals with the possibility of special attacks on the biometric system of authentication by face image. A new method of detecting fake attacks (spoofing attacks) is proposed. The method is based on the use of an artificial convolutional neural network which was trained using a ReplayAttack Database from Idiap Research Institute. The obtained results show high efficiency of the proposed method of detecting spoofing attacks: the probability that an attack will be detected is 94.98%. Keywords: Convolutional neural networks · Biometric authentication · Spoofing attack

1 Comparative Analysis of Biometric Authentication Today, biometric identity authentication technologies are widespread. These systems are implemented not only in enterprises, specialty facilities, but also on smartphones of ordinary users and in online applications. The problem of choosing one of the authentication methods remains urgent. Figure 1 shows the distribution of biometric technology in the global market. According to J’son & Partners Consulting, fingerprint scanning is the most popular way of © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 85–98, 2021. https://doi.org/10.1007/978-3-030-58359-0_6

86

A. Kuznetsov et al.

biometric authentication. It is used by 47% of companies. This is followed by facial recognition (32%), iris scanners (12%), vein pattern (4%) and voice recognition (6%).

Fig. 1. Use of biometric technologies in the global market as of 2019.

Figure 1 shows that one of the most common algorithms is fingerprint and face authentication. Following are face recognition technologies, iris, vein pattern and other techniques. Most often, comparative analysis of biometric authentication methods is performed based on first- and second-order errors. False Reject Rate (FRR) is the likelihood that an authentication system will not be able to identify a registered user (or is often said to accept “valid” as “imposter” user). A second kind of FAR (False Accept Rate) error is the probability that the system identifies an unregistered user (that is, accepts a “ imposter” for “ valid” user). The values of FRR and FAR are calculated by the formulas: FRR =

Q , N

(1)

FAR =

R , N

(2)

where Q is the total number of erroneous access denials, R is the total number of erroneously granted accesses, N is the total number of objects being tested [1]. The values of FAR and FRR for different biometric authentication methods are shown in Table 1. Based on the results of the comparative analysis, we conclude that one of the most accurate methods is biometric authentication on the iris. In this case, the FAR and FRR indicators are the lowest, thus achieving the highest accuracy and authentication. It should be noted that the use of the iris is not always a convenient way of authentication. It usually requires the use of specific hardware devices to accurately scan the iris. In addition, fixing the eye near the hardware scanner is sometimes impossible or

Biometric Authentication Using Convolutional Neural Networks

87

Table 1. The values of FAR and FRR for different methods of biometric authentication [2] Biometric authentication method FAR

FRR

Fingerprint

0,001%

0,6%

2D face recognition

0,1%

2,5%

3D face recognition

0,0005%

0,1%

Iris

0,00001% 0,016%

Vein pattern

0,0008%

0,01%

psychologically not perceived by the user. These circumstances diminish the practicality of this technology, as evidenced by the data in Fig. 1 on the diffusion of different biometric authentication technologies. Following in terms of performance are vein pattern technologies, face recognition, and more. Particularly noteworthy are face recognition technologies using 3D scanning. Recently, these techniques are gaining ground due to the considerable progress made in the creation of cheap and extremely high quality photo and video cameras. In particular, virtually all modern smartphones are equipped with powerful tools that can be used, including for the implementation of biometric authentication technologies for photo or video image of the face. Table 2 lists the advantages and disadvantages of this technology. Table 2. Advantages and disadvantages of the face image authentication algorithm Method Advantages

Disadvantages

2D

Low cost of implementation, high speed, Low statistical reliability; demanding availability of ready-made solutions and lighting; rejection of any external databases obstacles

3D

No contact with the scanner Low sensitivity to external factors High level of reliability and accuracy of scanning

Relatively expensive equipment, low number of databases (these shortcomings are quickly eliminated with the improvement of modern photo and video tools)

Therefore, in terms of convenience, prevalence and accuracy (see Tables 1 and 2), biometric face recognition authentication technologies are the most attractive. The main problem with the practical application of biometric authentication on a person’s face is the ability to perform special substitution attacks. These are so-called spoofing attacks designed to impersonate another person through fake biometric data. This paper examines this situation in detail and develops a new method for detecting counterfeit attacks by using an artificial convolutional neural network. Therefore, the proposed biometric system is reliable and secure, as well as convenient and unpretentious to human face scanning hardware.

88

A. Kuznetsov et al.

2 Biometric Authentication System and Its Vulnerabilities In general, biometric authentication system works in two modes: 1. In the registration mode, the biometric characteristic is obtained by means of a special sensor and recorded in the database; 2. In authentication mode, the biometric data obtained is used by the system to verify that the user for whom he or she is impersonating is. 3. In any case, any biometric system should consist of the following modules (Fig. 2): • a sensor module that receives biometric user data; • module for obtaining biometric characteristics, which is intended to extract the basic values of biometric characteristics; • a matching module that compares the biometric characteristics obtained with those stored in the database; • a decision module based on which a user’s identity is validated or denied.

Fig. 2. Modules of the biometric system of authentication (identification) of users

A biometric authentication system can be vulnerable to two types of errors: invasion and denial of service. Denial of service is such a faulty version of the system in which a legitimate user is recognized as an attacker. Intrusion is an erroneous version of a system in which an attacker is recognized as an authorized user. Such errors can be the result of both incorrect operation of the system itself and attacks by the attacker. A schematic of biometric authentication system errors and their possible causes is shown in Fig. 3. Incorrect operation of the biometric authentication system is associated with two types of errors - FAR and FRR. Their definitions and calculation formulas are presented in the first section. Ideally, a biometric authentication system would have FAR = 0 and FRR = 0. However, such values are in reality not achievable. The value of either of the two errors can be reduced. The disadvantage is that the second value will increase. Typically, the system settings are adjusted to achieve the required error rate, which determines the corresponding error rate. With these options, the attacker does not need to take any action. Such invasions are also called zero-force attacks.

Biometric Authentication Using Convolutional Neural Networks

89

Fig. 3. Biometric system vulnerabilities

Incorrect operation of the biometric authentication system may be related to attack by an attacker. They can be roughly divided into two types: external and human-related attacks. The first type of attack is that an attacker can gain access to the system by colluding, malpractice, etc. External attacks can be related to attacks on sensors, modules for obtaining and comparing biometric characteristics, intermodal connections, as well as attacks on the biometric sample database. Most attacks on the module for obtaining and comparing biometric characteristics (Trojans) and on intermodal connections (human attack in the middle and attack playback) are also applied to password authentication systems. Today, cryptographic measures to counter such attacks (timestamp, mutual authentication, etc.) exist and are successfully operating. They allow you to prevent or minimize the effects of such attacks. The most important types of attacks on the biometric system are attacks on the database of biometric templates and sensors to obtain biometric characteristics. The attack on the template base is that the biometric sample of the user can be made available to the attacker. This is a serious enough problem because if the attacker receives biometric parameter data, authentication with the biometric parameter will not be possible in the future. The solution to this question can be achieved by storing in the database not the biometric samples themselves, but the key formed on the basis of biometric data. Such a crypto system is called cryptosystem with key generation. In a biometric cryptosystem with key generation, a pseudo-random sequence (key) is generated directly from the biometric data of the user, which is not stored in the system. A key generation system based on fuzzy extractors is described in [3]. Attacking biometric sensors or Spoofing attack is a case where a person or program disguises itself against another by falsifying its biometric data at the input of the system,

90

A. Kuznetsov et al.

thereby gaining an illegal advantage in the future. The attack is as follows: a fake biometric feature of a person, such as a face photo, a plasticine finger, a fake eyeball, and so on, is submitted to the biometric authentication system sensor. If these fakes are of the appropriate quality, the system may confuse them with true (real) features and grant access to the attacker. Today, there are methods that help counteract a spoofing attack. Such methods can be divided into two classes: active and passive. Active methods are intended to determine the characteristics inherent in a living person. An algorithm can analyze eye movements, require a smile, wink, and more to authenticate user by face. Passive methods use only one image for analysis and do not require specific actions from the user. The use of the latter is more acceptable, since it does not at first require additional user action, and secondly it does not allow the photo to be displayed in the interval between the definition of vitality and authentication. In recent years, deep neural networks have been actively used to counteract spoofing attacks. Thus, in [4], a system for detecting replay attacks was presented, which was presented at the ASVspoof 2017 competition for this problem. The best neural network based system showed an EER error of 6.73% on a subset of unknown attacks, which is 72% better than the basic method presented in the competition. In [5] uses a AlexNet-type neural network to detect a spoofing attack, and investigates the effect of image size on the quality of such a network. In [6], a combination of a convolutional neural network and a recurrent LSTM type network was used to classify frames. Despite the development of this area, further work remains relevant. High pixel density, natural color rendering makes biometric images close to real biometric data. That is why the task of developing new algorithms for detecting spoofing attacks and upgrading existing ones remains relevant. The following describes a new method for detecting fake spoof attacks by using an artificial convolutional neural network.

3 Face Recognition System Using a Convolutional Neural Network Today, the use of neural networks is a common practice in pattern recognition tasks. Modern libraries with standard feature sets help greatly simplify the development of new learning algorithms and data processing. A convolutional neural network based authentication system consists of the following steps: 1. User takes a photo of the face in real time. 2. A trained convolutional neural network finds faces in an image. 3. Unique features are removed from the face image found using another trained convolutional neural network [7]. 4. The unique features of the user are compared to those stored in the database. 5. The decision module calculates the difference between the unique attributes of the input image and those stored in the database. Based on a specially set threshold, a decision is made as to whether or not a given user is allowed to access the system.

Biometric Authentication Using Convolutional Neural Networks

91

Such a biometric authentication system is vulnerable to spoofing attacks [8]. We offer a new method of recognition of spoofing attack, which is based on the use of deep convolutional neural network. In the general case, the architecture of the convolutional neural network for face recognition in the image can be represented as follows, as shown in Fig. 4 [9]. This artificial neural network architecture was named because of the convolution operation, the essence of which is that each fragment of the image is multiplied by the convolution matrix (nucleus) element by element, and the result is summed up and written to a similar position of the original image. Such a convolutional neural network includes levels of detection of characteristics in a face image, each alternating with a pooling layer to obtain features that are inherent in different image variants (distortions, rotations at different angles, etc.). The model in question can also find a smiley face, finding the difference between a normal image and an image fed into the network input.

Fig. 4. A convolutional neural network architecture for face recognition in an image

The YOLOv3 deep convolutional neural network was used as the network used to find the face in the photograph [10]. YOLO (You Only Look Once) – the convolutional neural network architecture quite popular at the moment. It is used with great success in the object recognition task in the image. This network allows you to find objects in the image in real time and it is the state-of-the-art network for today. The YOLOv3 neural network at a frame rate of 30 indicates an object recognition accuracy of 57.9% on COCO test-dev [11], which is quite high. YOLOv3 is an advanced version of the YOLO network. It consists of 106 convolutional levels and better detects the image in the photos than its previous versions.

92

A. Kuznetsov et al.

A distinctive feature of the YOLOv3 architecture is the presence of three layers at the output of the network, each of which can find objects of different sizes in the image. A schematic representation of such a network is shown in Fig. 5.

Fig. 5. Convolutional neural network architecture YOLOv3

The YOLOv3 architecture was implemented using the Darknet framework [12]. The result of the work at this stage will be allocated face images in the photo (Fig. 6).

Fig. 6. The results of the convolution neural network YOLOv3

4 The Proposed Method of Detecting Spoofing Attacks After the face has been found in the image, you have to make a decision whether or not it is an attempt at a spoofing attack. The proposed method of detecting fake attacks at the entrance of the deep convolutional network is a face image of 128 × 128 pixels, and the output has a binary value of 0 or 1: • 1 (or “yes”) when the input is recognized as a spoofing attack; • 0 (or “no”) - when the input data is a real image of a person’s face.

Biometric Authentication Using Convolutional Neural Networks

93

To train such a network, a reverse error propagation method was used, which was to propagate error signals from the neural network outputs to its inputs. This method is used to minimize the multilayer perceptron error and obtain the desired output at the network output. During convolution training, the method is used when updating the weights of the multilayer perceptron [13, 14]. Data from the Idiap Research Institute contained in the Replay-Attack database were used to train our convolutional neural network [15]. The Replay-Attack database combines 1,300 video clips of photos and video attacks for 50 different people under different lighting conditions. Data from this database is already divided into 4 subgroups: • Train data (“train”) to be used to train a convolutional neural network against counterfeit attacks; • Validation data (“valid”) used to find the optimal threshold for image classification; • Test data (“test”) - data that will measure the effectiveness of the developed method; • Enrollment data (“enroll”) - data that can be used to validate counterfeit sensitivity in facial image detection algorithms. Data that appears in one of the subgroups (train, valid, test, enroll) does not appear in any other set. The extension of our data was done using the image augmentation algorithm: • Adding a Gaussian noise with a value scale = 0.05 ∗ 255 (standard deviation from the normal noise-generating distribution). • Adding a Gaussian Blur with a value sigma = 0.5 (standard deviation of a Gaussian distribution). • Turns at angles from −45 to 45°; • Enlarge or reduce face images. In the next phase of the study, it was necessary to justify specific metrics to measure the effectiveness of the method developed. Our convolutional output neural network can accept two possible binary values 1 and 0 (yes/no). Therefore, we have the following set of possible situations (Table 3). Let us define the null hypothesis H0 for our network as an attempt at a spoofing attack. Then, the alternative hypothesis H1 will be defined as the input image to our non-spoofing recognition algorithm. For each image at the input of our algorithm we will give possible solutions or the conclusion that before the image will be recognized as a spoofing attack or not. Table 3. Possible situations when fake attacks are detected The real value Model predicted result

H0

H1

H0

True positive

False negative

H1

False positive

True negative

94

A. Kuznetsov et al.

For all input features, the following rule was adopted: if the value of the output of the convolutional network algorithm is less than 0.5, then the conclusion is accepted that the image is real, otherwise the conclusion is taken that the image is an attempt at a spoofing attack. Table 3 gives the following definitions: • • • •

True positive - the network correctly recognized the spoofing attack; False positive - the network recognized the spoofing attack as a real face image; True negative - the network correctly recognized the real face image; False negative - the network recognized the real face image as a spoofing attack.

Our system will be configured to minimize first-order error, as a situation where a model recognizes a spoof attack as a real face image can have critical consequences. Next, we need to test the effectiveness of our system, because we need to know how the model will behave when dealing with data that was not in the training sample. This is done by taking a set of images that the algorithm has not yet seen. In our case, this is data from the “test” set. During testing, all results are stored in a special form. After testing the entire set, we get four values (True positive, True negative, False positive, False negative.), which are the results of the model. To better evaluate the performance of our model, we use Precision and Recall metrics. Precision shows the proportion of detected attacks from all attacks that were on the dataset, and Recall shows the proportion of correctly recognized attacks from all images that were recognized as attacks [16]. Indicators are calculated using formulas 3 and 4: Precision = Recall =

True positives , True positives + False positives True positives , True positives + False negatives

(3) (4)

where: True positives – total number of correctly recognized attacks; False positives – total number of unrecognized attacks; False negatives – total number of incorrectly recognized attacks. The selection of such indicators is due to the need to minimize the first kind of error. The architecture of the developed convolutional neural network is shown in Fig. 7. So our neural network has 5 convolutional levels consisting of Convolutional (Conv2D) layer and MaxPooling2D layer. The last digit in the Output Shape vector is the number of filters in the convolutional layer. The filter size in all convolutional layers is 3 × 3 and the filter size in the MaxPooling2D layer is 2 × 2. After the last convolutional block, there is a Flatten layer that changes the dimension of the 3-D matrix into a 1-D vector. The following are two Dense layers, the first one being the size of 1024 neurons with relu activation function, the second one with 1 neuron and sigmoidal activation function to obtain the probability of detecting a substitution attack.

Biometric Authentication Using Convolutional Neural Networks

95

Fig. 7. Architecture of the developed convolutional neural network.

The neural network weight optimization was performed using the Adam algorithm [17]. The input datas to the algorithm assumed the following values: • α = 0.001: learning speed; • ε = 10−8 ; • β1 , β2 ∈ [0, 1]: Exponential decay rates for moment estimates (β1 = 0.9, β2 = 0.999); • f (θ ): Stochastic function with parameter θ ; • θ0 : Vector of initial characteristics. At the output of the algorithm optimization of weights we get θt (The resulting parameters). Training neural network was held with the following values of parameters: • • • •

learning_rate (a training step or the rate at which weights change) - 0.001; epochs (the number of training cycles) - 25; batch_size (the number of images that are fed to the model at the same time) - 32; loss (a special function that is used to optimize the neural network): binary_crossentropy.

Our goal was to minimize the loss function by optimizing its parameters (weights). Loss is calculated using a special function by comparing the real and predicted values of the neural network. We use the Adam gradient descent method to adjust neural network weights to minimize loss. loss is responsible for exploring the convolutional neural network. Since in our case there is a problem of binary classification at the output of the network, the loss function will take the value binary_crossentropy loss [18, 19]. It is defined as follows: Hp (q) = −

N 1  yi · log(yi ) + (1 − yi ) · log(1 − yi ), N 

i=1



(5)

96

A. Kuznetsov et al. 

where: yi - this is a real value (1 or 0); yi - is the value predicted by the neural network. That is, binary_crossentropy loss measures the performance of our model, the output of which is a probability value between 0 and 1. Loss increases when the predicted probability deviates from the real value. The ideal model should have a loss value close to zero. All software was written in Python programming language using such frameworks: tensorflow [20], opencv [21], imgaug [22]. According to the results of the experiment, the value of the number of epochs was set, exactly 20. The loss function of our network takes the value 0.18, and the overall accuracy on the “valid” data set is 0.95. Figure 8 shows how our model changes in training and validation data.

Fig. 8. Change the accuracy of the model during training on validation and training data.

To increase the efficiency of our model, we need to find the best decision threshold, which is 0.5 by default. The intersection point of the FPR and FNR probabilities was found, which was determined to be the optimal threshold. Figure 8 shows how FPRs and FNRs change with the decision threshold being changed.

Fig. 9. Graph of the false positive rate and false negative rate of change threshold decision

Biometric Authentication Using Convolutional Neural Networks

97

As can be seen from Fig. 9, the optimal decision threshold should be up to 0.59 in order to minimize errors of the 1st and 2nd kind. Since the false negative rate does not increase much with the increase of the threshold, while the false positive rate decreases, it was decided to set a decision threshold of 0.7. Once the optimal decision threshold has been found, we need to measure the performance of our model on new data. In our case, such data is “test” data from the database in use. The results of the algorithm using different datasets are shown in Table 4. Table 4. The results of the algorithm work on different datasets Training database

Validation database

Test database

Precision, %

0.98

0.95

0.95

Recall, %

0.96

0.91

0.92

The values obtained in the table were calculated on a trained neural network. The probability of a spoofing attack being detected is 94.98%.

5 Conclusions In the course of the work, a method was developed for determining attacks of fake biometric facial images based on the use of an artificial convolution neural network. The developed method was proved itself well on all datasets from the Replay-Attack database. On the test dataset Precision is 0.95% and Recall is 0.92%. The probability that a spoofing attack will be detected by a neural network is equal 94.98%. Further research will be to test the method on real user data and adjust the settings of a developed convolutional neural network accordingly. These studies will also be useful for improving other methods of biometric authentication, in particular, based on the use of fingerprints [23, 24].

References 1. Tsarov, R.Yu., Lemeha, T.M.: Biometric technology, Odessa (2016). 140 p. 2. Face recognition technology - a new era in video analytics, video surveillance and access control systems. https://securityrussia.com/blog/face-recognition.html 3. Lutsenko, M., Kuznetsov, A., Gorbenko, Y.: Key generation from biometric data of Iris. In: The Fourth International Conference on Information and Telecommunication Technologies and Radio Electronics, UkrMiCo 2019, Odessa (2019) 4. Lutsenko, M., Kuznetsov, A., Gorbenko, Y., Oleshko, I., Pronchakov, Y., Kotukh, Y.: Key generation from biometric data of Iris. In: 2019 International Conference on Information and Telecommunication Technologies and Radio Electronics (UkrMiCo), Odessa (2019 in press) 5. Lavrentieva, G.M., Novoselova, S.A., Kozlova, A.V., et al.: Detection methods for spoofing attacks of repeated reproduction on voice biometric systems. Sci. Tech. J. Inf. Technol. Mech. Opt. 18(3), 428–436 (2018)

98

A. Kuznetsov et al.

6. Yang, J.: Learn convolutional neural network for face anti-spoofing. arXiv preprint arXiv: 1408.5601 (2014) 7. Xu, Z.: Learning temporal features using the LSTM-CNN architecture for face anti-spoofing. In: ACPR, pp. 41–45 (2015) 8. FaceNet: A Unified Embedding for Face Recognition and Clustering. https://arxiv.org/pdf/ 1503.03832.pdf 9. Lazarick, R.: Spoofs, subversion and suspicion: terms and concepts. In: Proceedings of the NIST International Biometric Performance Conference (IBPC) (2012) 10. Matsugu, M., Mori, K., Mitari, Y., Kaneda, Y.: Subject independent facial expression recognition with robust face detection using a convolutional neural network. Neural Netw. 16(5–6), 555–559 (2003) 11. Redmon, J., Farhadi, A.: YOLOv3: An Incremental Improvement. University of Washington. https://arxiv.org/abs/1804.02767 12. YOLO: Real-Time Object Detection. https://pjreddie.com/darknet/yolo/ 13. Darknet. https://github.com/pjreddie/darknet 14. Rumelhart, D.E., Hinton, G.E., Williams, R.J.: Learning Internal Representations by Error Propagation, 1 September 1985. https://doi.org/10.21236/ada164453 15. LeCun, Y., Bengio, Y., Hinton, G.: Deep learning. Nature 521(7553), 436–444 (2015). https:// doi.org/10.1038/nature14539 16. Chingovska, I., Anjos, A., Marcel, S.: On the effectiveness of local binary patterns in face anti-spoofing. In: BIOSIG-2012. IEEE BIOSIG (2012) 17. Ting, K.M.: Precision and recall. In: Sammut, C., Webb, G.I. (eds.) Encyclopedia of Machine Learning. Springer, Boston (2011) 18. Kingma, D.P., Ba, J.: Adam: A Method for Stochastic Optimization. Submitted on 22 Dec 2014 (v1), last revised 30 Jan 2017 (this version, v9). https://arxiv.org/abs/1412.6980 19. Janocha, K., Czarnecki, W.M.: On Loss Functions for Deep Neural Networks in Classification. Faculty of Mathematics and Computer Science, Jagiellonian University, Krakow, Poland (2017) 20. Abadi, M., Agarwal, A., Barham, P., et al.: TensorFlow: large-scale machine learning on heterogeneous systems (2015). Software available from tensorflow.org 21. Bradski, G.: The OpenCV library, Dr. Dobb’s J. Softw. Tools 25, 120–125 (2000) 22. Jung, A.B.: imgaug (2018). https://github.com/aleju/imgaug 23. Kuznetsov, A., Kiyan, A., Uvarova, A., Serhiienko, R., Smirnov, V.: New code based fuzzy extractor for biometric cryptography. In: 2018 International Scientific-Practical Conference Problems of Infocommunications. Science and Technology (PIC S&T), Kharkiv, Ukraine, pp. 119–124 (2018). https://doi.org/10.1109/INFOCOMMST.2018.8632040 24. Rassomakhin, S., Kuznetsov, A., Shlokin, V., Belozertsev, I., Serhiienko, R.: Mathematical model for the probabilistic minutia distribution in biometric fingerprint images. In: 2018 IEEE Second International Conference on Data Stream Mining & Processing (DSMP), Lviv, Ukraine, pp. 514–518 (2018). https://doi.org/10.1109/dsmp.2018.8478496

Transdisciplinary Fundamentals of Information-Analytical Activity Stanislav Dovgyi(B)

and Oleksandr Stryzhak(B)

National Center “Junior Academy of Sciences of Ukraine”, Degtyarivska St. 38-44, 04119 Kyiv, Ukraine [email protected], [email protected]

Abstract. The information and analytical activities of experts have one of the main problems which are the unstructured information resources usage to be displayed in various network documents and represent a passive distributed system of knowledge in fact. This research considers transdisciplinary of such information resources as a metacategory that takes into account the hyperproperties of Big Data (Big Data), namely: a) reflection which implements the principles of integration, the consistency and their behavior integrity and guarantee; b) recursion which implements the recurrence category of their operational transformation forms during activation; c) reduction on the basis of which these forms decomposition principle is realized. Their interpretation in the case of Big Data processing is implemented in the following areas: i) the information resources structural analysis; ii) forms of interaction with information resources; iii) definition of the mechanisms for identifying criteria for selecting appropriate contexts that needed for the expert analysis. The actuality of the such procedures implementation is based on the need to create the conditions for supporting the effective a large number of diverse information arrays processing for the information and analytical activities of experts. This understanding for solving the Big Data processing problem is supported by the implementation of component architecture of the services to support the analytical processes of the experts from various thematic sphere of activity. Keywords: Transdisciplinarity · Ontology · Narrative · Discourse · Taxonomy · Cognitive system

1 Introduction 1.1 A Subsection Sample Information and analytical activities of experts in the modern information space is network-centric and is implemented via transdisciplinary interaction of all its information resources and processes [1–7]. The presence of transdisciplinary interconnected processes of production, processing, storage, dissemination, and use of large amounts of information and knowledge in the form of unstructured documents is one of the main © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 99–126, 2021. https://doi.org/10.1007/978-3-030-58359-0_7

100

S. Dovgyi and O. Stryzhak

factors of these processes [8–10]. However, it is necessary to ensure the objectivity of the analytical conclusions and decisions made based on the analysis of these documents. And it should be borne in mind that these passive knowledge systems reflect different domains, have a significant number of interdisciplinary relationships, and are based on the use of different information technologies and standards. The list of network passive knowledge systems, which is quite extensive in terms of volume and topics, includes such categorical concepts as dictionaries, thesauri, taxonomies, narratives, narrative discourses, linguistic corpora, etc. [1, 2, 11–14]. All of them are classified as poorly structured, and belong to Big Data according to the totality and nature of the presentation [1, 2, 15, 16], and are also characterized by multifaceted and multiple latent connections, etc. We introduce the following concepts: 1) Cognitive system, the cognitive structure is a system of human cognition, which occurs in thinking as a result of their character’s development, education, training, observation and reflection on the world, and in terms of information technology means the stable definition of goals patterns, decision-making algorithms according to a certain state of the objects of analysis in order to obtain reliable results in the future. 2) The narrative is a universal model of establishing content links between the texts of documents, sites, and other forms of information presentation, which considers the process of linking them as a form of presenting content links as graphs. 3) Discourse is reasoning, proof, defined in terms of information technology as certain algorithms and software services based on them. 4) Taxonomy is a structure, order, regularity, based on the principles and practical application of algorithms for classification, systematization of complex, hierarchically connected objects, or entities. The transformation of such passive knowledge systems reflected in the form of documents that are formed and reflect the descriptions of certain processes and their properties is a very important issue. However, it requires the implementation of procedures for their transformation, at least into an interactive form, which determines the conditions for the implementation of interaction with these systems of active knowledge of relevant experts. The processes of structure solution of these problems are intellectual and are based on solving the following categories of meta-tasks i.e. structuring; problem analysis/selection; synthesis; choice. The interaction of experts and specialists through the meaningful discourse [6, 17–20] with passive knowledge systems is realized with them. It provides transdisciplinary transformations of all documents in an interactive form [1, 2, 19].

2 Transdisciplinary Features of the Expert’s Activity in the Information Environment As we have already noted, the information and analytical use of online documents require their transformation into an active form of knowledge. And if the active form of knowledge can be represented primarily as an ontological system [1, 2, 5, 6, 12, 17–21], represented as: OI = X, R, F, A, D, Rs 

(1)

Transdisciplinary Fundamentals of Information-Analytical Activity

101

where X – is the set of concepts of a given domain; R – is a finite set of semantically significant relations between the domain concepts. F – is a finite set of interpretation functions given on concepts and/ or relations. A – is a finite set of axioms used to write always true statements (definitions and limits) in terms of the domain subject; D – is a set of additional definitions of concepts in terms of the domain subject; Rs – is a set of limits defining the scope of conceptual structures of a particular domain subject. Then passive knowledge systems are certain texts, reflected as a consistent presentation of certain concepts, stable conditions for their existence, descriptions of their properties and functionalities, and so on. That is, we have a definition of ontology (1) in some following system: Onr =

(2)

where Onr is a document with sequentially defined descriptions of K contexts (description) of X concepts, and p is a strict order relationship that determines the conditions of existence of X concepts in the text. Then we will define Onr as a narrative, determined by the strictly consistent presentation of descriptions of certain facts, processes, and their properties [1, 2, 6, 17–21]. So, a narrative is an ontology of the form (2), which is a simple text that passively presents a certain system of knowledge. A certain non-empty set of such narratives forms a passive transdisciplinary knowledge system in the form of a set of multidisciplinary documents. Their coherence is possible and the formation of certain new documents on their basis is made based on transdisciplinarity. When considering systems of the form (2), we can establish only strict intercontextual links. That is to form the final strict sequences: −→ −−→

(3)

Expression (3) interprets the passive integration of narrative forms (2). To transform it into an interactive form (1) it is necessary to determine the conditions of integrated operability [22–29], namely: analysis and structuring (processing), synthesis (communication), and choice (decision making) [1, 24, 30–34]. Based on the fact that the operability of a random text [22–29] (2) and (3) is determined by the following hyperproperties: Rf - reflection, Rk - recursion and Rd - reduction. These hyperproperties form a closed set R3, which provides the binding and dynamic change of ordering of contextual descriptions of the narrative Onr [2, 30]. These hyperproperties are transdisciplinary and interpretive [1–4, 35, 36].   R3 = Rf , Rk , Rd (4) where Rf reflection - implements the categories of integration, system, and absence of meaningful gaps in the arrays of Big Data and provides tracking and modification of their own structures and the stability of their behaviour during activation;

102

S. Dovgyi and O. Stryzhak

Rk recursion - implements the category of processes, phenomena, and forms of their operational reflection recurrence during their activation in the information space Rd reduction - implements the methodological principle, and complex phenomena can be decomposed and fully explained according to it based on laws of simpler processes and objects. The cognitive nature of these hyperproperties is realized in their interpretation by hyperfunctions such as Big Data analysis, its structuring, synthesis, selection, and decision-making, which are cognitive [1, 26–29, 34–37]. Based on certain concepts such as ontology, narrative, reflection, recursion, reduction, etc., we define the category of transdisciplinarity. Since we consider the process of information and analytical activities of the expert, we provide a narrow definition. Under the transdisciplinarity of certain knowledge systems, we understand the conditions of application of reflexively active recursive reduction FR3 . Its properties will be discussed below. It means that we can extend the ontology (2) by the R3 set and the hyperfunction of its interpretation FR3 Onr = → Ond =

(5)

As we see through transformation (5), by substituting the set of hyperproperties R3 and the functional extension FR3 , instead of the relationship of strict order p we obtained the ontological system Ond , which provides different types of connections between X concepts and K. We will describe the rules of forming such connections below. A general scheme of transformations of network documents into an ontological system is shown in Fig. 1.

Fig. 1. Organization of interaction with network documents based on a transdisciplinarity.

Transdisciplinary Fundamentals of Information-Analytical Activity

103

As we see from Fig. 1. the transdisciplinary nature of the interaction of experts with network documents is made by their transformation into interactive ontological systems. They form narrative discourse, which implements intercontextual coherence based on hyperproperties of reflection, recursion, reduction, and cognitive functions: analysis, structuring, choice etc. Cognitive tools ensure the use of all the contextual coherence of the integrated narrative of the information space of information and analytical activities of experts. The interaction of experts and specialists with an integrated narrative of descriptions of various knowledge systems is based on cognitive tools that provide transdisciplinary transformations of all the documents into an interactive form. Moreover, these cognitive services should be focused on processing the integrated information generated in interaction with information resources based on intercontextual connections. The multiplicity of these intercontextual connections can be represented as the taxonomies [2, 13, 14], which in turn can be represented as mathematical statements [27, 30, 38–40]. Then we can represent the ontological system Ond , in the right part of the expression (5), as individual cases: Od =

(6)

Od =

(7)

Od =

(8)

All these cases (6)–(8) represent a category of discourse between documents, as a cognitive and communicative act, which simultaneously implements an integrated description of selected processes and their interpretation as reflection and representation based on binary intercontextual communication [7, 10, 17–20]. One of the properties of discourse is the ability to reflect the coherence of two or more narratives. Then the narrative discourse [17–20, 26, 29] will be defined as the interaction of documents presented as the ontological systems based on verbally active hyperfunction FR3 , which implements their systematization, i.e. analysis, structuring, classification, criteria, synthesis, and evaluation, etc. [2–4, 10, 14, 29, 35–37]. The FR3 function is verbally active if and only if the ontological system it interprets includes the whole set R3. So, the narrative discourse of a set of network documents will be presented as a right part of the expression (5): Ond = .

(9)

3 Taxonomy as a System of Structural Analysis of the Narrative The definition of narrative, discourse, and narrative discourse allows us to consider the category of a unified information space based on the integration of transdisciplinary

104

S. Dovgyi and O. Stryzhak

ontologies [1, 2, 7, 26, 41–44]. However, a more detailed consideration of the process of interaction with network information resources requires the selection of a category for which the structural representation is a unary property [2, 30, 40–45]. Taxonomy is such a category [1, 2, 7, 12–14, 26, 45]. The structural display of both the separate document and rather big on volume collection of documents is made based on it. In this case, the digital collection of documents means the procedure of systematization of documentary network resources (Big data sources) by a set of natural language texts, combined by one feature or set of features (linguistic, conceptual, pragmatic, temporal, stylistic, functional, intentional, etc.). Digital collections of documents create conditions for linguistic and semantic analysis of texts, which allows you to find phrases that use terms in the texts of relevant documents automatically. The taxonomy of documents (T ) is considered as a certain result of the application of the cognitive procedure of structuring of text arrays based on a hierarchical and systematological representation of their terminological system. The result of the application of the procedure of taxonomization of texts is the representation of their structure as a graph without cycles [40, 44], where each point contains the relevant contexts. Its content consists of semantic descriptions and characteristics of relevant terms and phrases. The taxonomy provides the selection of classification units of the text array, which characterize its semantics and purpose, as well as reflects the order of interaction between terminological constructions. Taxonomic representation of a certain set of documents characterizing the various processes creates technical conditions for making of their digital collection. Note that random taxonomy, as a multiple hierarchical ordering of the terms of a particular collection of documents, can be represented as a growing pyramidal network [44]. Classically, according to the definition proposed by V. Gladun [44, 45], the pyramidal network Ψ - means an acyclic  oriented graph Ψ = (X , E), where there are no points with one incoming edge. X = xi |i =|1, n1 is a set of network points (document concepts), where xi is a random network point, n1 is the number of points in the network. E =  ei |i =|1, n2 is the set of network edges, where ei is a random edge, n2 is the number of edges in the network. An example of a semantic network of pyramidal structure proposed by V. Gladun [14, 66] is shown in Fig. 2. Points with no arcs form terminal concepts (terminals), other points form classes of concepts (classes). Terminals correspond to separate values of signs from objects descriptions of. Classes correspond to combinations of attribute values generally identifying the object, or to the corresponding common parts of the descriptions of several objects. The set of points of a pyramidal network is a set of X concepts of a pyramidal network, where each concept xj represents a certain linguistic lexical unit: xi = concept |word |phrase

(10)

A subgraph of a pyramidal network that includes a certain point C ∈ X and all points from which there is a path to it is called a pyramid of point C. The points of the pyramid of point C form its subset, which we note as XC . Points of XC and directly connected to the C edge will be called the 0-subset of this point. The set of points including a point C is called its superset noted as XC [24, 28].

Transdisciplinary Fundamentals of Information-Analytical Activity

105

Fig. 2. An example of a semantic network of pyramidal structure, (as defined by V. Gladun)

It is clear that XC ∪ XC = ΨC , where ΨC – is a pyramidal network formed by all concepts that have common edges with point C. Such taxonomic structures in form of a growing pyramidal network implement a transdisciplinary categorization of contexts, as a systemic, dynamic formation of classes of contextual descriptions based on the formation of stable binary connections between certain terms, phrases, and word forms. Mapping the semantic coherence of concepts and their contexts, reflecting the relevant information resources in form of a pyramidal network Ψ , is a necessary condition for the formation of many taxonomies that can reflect the structural diversity of the whole narrative of all text arrays that form these resources. These pyramidal networks Ψ are univalent [57, 58] to taxonomic structures T: Ψ ∼ =T

(11)

Then there are always non-empty sets of taxonomies and pyramidal networks of a certain narrative, which are also univalent to each other. Such structural mappings can be called narrative taxonomic diversity (NTD): ({Tj |j = 1, m} ∼ = {Ψi |i = 1, n}|j ≥ i, m ≥ n)

(12)

So, random taxonomy T can be formed based on a certain finite set of pyramidal networks as {Ψi |i = 1, n}, which we call narrative taxonomic diversity, and note as Tnd : n Tj ∼ Ψi (13) = i=1

Expressions (10)–(13) reflect the fact that a certain taxonomy T is formed by sets of concepts that make a nonempty set of pyramidal networks Ψ = (X , E) of a certain

106

S. Dovgyi and O. Stryzhak

narrative. Thus the {Ψi |i = 1, n}, set, which is determined by the set of X concepts, is a semantic structural reflection of the narrative representation of information resources. Based on this, it can be argued that the X concepts of narrative taxonomic diversity (13) define some terminological field TF of this narrative, and this terminological field is open. That is, a random taxonomy T can be supplemented by the newest pyramidal networks (12), which have additional concepts x k | k > n. The set of edges E of a pyramidal network Ψ , determines only the presence of binary connectivity between its concepts. The orientation of these edges allows you to group concepts into such categories as terminal, class, subset, superset, concept. The terminal concept x t has no input arrows, i.e. it is not commutative compared to the binary edge ei : (xt ei xi ) = (xi ei xt )

(14)

X KL class defines all concepts that belong to a certain categorical non-terminal xi concept: X KL = {xi ei X KL |}

(15)

The subset X sb defines concepts for which the specified (basic) xi concept is categorical, and this concept can’t be terminal:     j j sb x j sb esb xi eisb ei+1 e(i+2) x(i+k) | i = 1, m , sb (i+1) (i+k−1)     Xi = , (16) k = i, m , j = 1, l eisb is an edge marked by an arrow entering the basic concept xi . The superset defines concepts that are categorical for the specified (basic) concept, and can be terminal:     j j sub x j sub esub xi eisub ei+1 e(i+2) x(i+k) | i = 1, m , sub (i+1) (i+k−1)     Xi = , (17) k = i, m , j = 1, l sup

is an edge marked by an arrow outcoming from the basic xi concept xi . sup, As can be seen from expressions (16) and (17), the edges ei , eisb the edges have the opposite orientation. According to this, we can determine the following: sup ei + eisb = el where el – Eulerian subgraph, for strongly connected concepts. sup ei × eisb = EL where El – is the set of Eulerian paths in the pyramidal network Ψi . The concept defines all categorical and terminal concepts from the specified x i at a certain distance. Its value is an integer and equal to the number of binary edges that connect them: ( n1 ei )|e = 1.

Xikc = {xi × l1 n1 ei xj } where l – is the distance from the xi concept to xj , n – is the number of binary edges from each concept. Based on the fact that the pyramidal network Ψi , is by definition a digraph without loops, there are no Eulerian subgraphs. However, the artificial formation of a nonempty set of Eulerian graphs and Eulerian paths based on the application of reflexively active recursive reduction FR3 is possible.

ei

Transdisciplinary Fundamentals of Information-Analytical Activity

107

The defined taxonomic functionality provides the technological basis for the effective processing of large amounts of information resources, inherent in modern information space. All documents that make it up are network-based. Therefore, taxonomy is the technological basis for ensuring the process of categorization of contexts of all information resources based on the systematological, dynamic formation of contextual descriptions classes that make up the entire narrative of these resources. However, it requires stable binary combinations between terms, words, phrases, and word forms. These combinations, in addition to mapping taxonomic structures (10)–(17) should implement connectivity between all system components of taxonomic structures and ensure their ordering in analytical research. For this, we introduce a group of hyperproperties that provide binding and dynamic reordering of contextual descriptions of the information resources narrative. This can be achieved through reflection Rf , recursion Rk and reduction Rd [2, 30, 39, 40]. These hyperproperties are transdisciplinary and cognitive [1–4, 35, 36] and form a closed R3 set:   R3 = Rf , Rk , Rd (18) The application of the hyperproperties of this set creates the conditions for the application of the ontological approach in the processes of analytical research of network weakly structured or unstructured information resources [1, 2, 7, 14–16]. Any taxonomic system is based on the structured representation of the subject area of its application. Usually, the classes of objects are the base of structuring [45–47]. And a set of all concepts, properties of which determine the semantics of the subject area are conditionally divided into these classes. The properties of concepts themselves allow us to determine the set of classes of relations between them due to the formation of binary relations with the hyperproperties of the set R3. Note that any ontological system is based on a structured representation of the subject area of its application. Usually, the classes of objects are the base of structuring [56], which conditionally divide a set of all concepts, the properties of which determine the semantics of the subject area. The properties of concepts themselves allow us to determine the set of classes of relations between them due to the formation of binary relations with the hyperproperties of the R3 set. The formation of each class is implemented as follows. Initially, each class is named and then filled with document concepts that have a binary relation with the class name based on the formation of a stable binary relation between its property and one of the hyperrelations of a closed R3 set. This relation looks like: t t , Rg ,>|rkl ∈ RTkl Rg ∈ R3, g ∈ {f , c, d } 1 in accordance with the multipath routing strategy adopted in the model. At the same time, for the transit nodes in the network (Ri = sk , dk ), the following restrictions are introduced in the basic model [19]: ⎧ k ≤ 1, k ∈ K; ai,j ⎪ ⎪ ⎪ ⎪ j:E ∈E i,j ⎪ ⎨ k aj,i ≤ 1, k ∈ K; (4) j:Ej,i ∈E ⎪ ⎪ ⎪ k − k = 0, k ∈ K. ⎪ ai,j aj,i ⎪ ⎩ j:Ei,j ∈E

j:Ej,i ∈E

3 Mathematical Model for Calculating the Maximum Number of Disjoint Paths with Secure Routing As an optimality criterion for solving the problem of calculating the maximum number of disjoint paths for secure routing the maximum of the following objective function has been chosen:  k wi,j ai,j . (5) J1 = Mk − Ei,j ∈E

194

O. Lemeshko et al.

In the objective function (5), the weighting coefficients wi,j are selected so that the choice of the set of disjoint paths Mk is oriented at minimizing the probability of compromise of the multipath as a whole [19] wi,j = − log10 (1 − pi,j )

(6)

where pi,j is the probability of compromising the Ei,j ∈ E link. In turn, the probability of compromising the entire path that is the part of multipath is calculated as

pn = 1 − (1 − pi,j ), (7) Ei,j ∈Ln

where Ln is an ordered set of network links making up the n th path. Whereas the probability of compromising a multipath is defined as: k

k PMP

=

M

pn .

(8)

i=1

Thus, the choice of the wi,j is based on the use of expression (6) and focused on the inclusion in a set of disjoint paths of links with a minimum probability of compromise. The introduction of the logarithm operation in (6) is dictated by the fact that when calculating the probability of compromising the paths (7), the corresponding probabilities of compromising the links obtained during the solution are multiplied, and the second term in (5) is an additive form. Solving the problem of calculating the maximum number of disjoint paths has been reduced to solving the optimization problem of Integer Linear Programming (ILP) with criterion (5) and linear constraints (1)–(4), since the control variables are Boolean, and the variables characterizing the number of disjoint paths used in process of secure routing (Mk ) take only integer values. R2 R5

R1

R7

R3 E1,3 R6 E1,4

R4

Fig. 2. The network structure for investigation the model for calculating the maximum number of disjoint paths with secure routing.

System of Solutions the Maximum Number of Disjoint Paths Computation

195

The features of the proposed model of calculating the maximum number of disjoint paths in secure routing will be demonstrated in the following example. The structure of the analyzed network is shown in Fig. 2, which consists of seven routers and nine links. Let the first router be the source node, and the seventh is the destination node. Consider, for example, two variants of forming a set of disjoint paths when applying the proposed model for the initial data presented in Table 1. Table 1. Initial data for investigation. Link

E 1,2 E 1,3 E 1,4 E 2,5 E 3,5 E 3,6 E 4,6 E 5,7 E 6,7

Probability of compromising the link Case 1 0.3

0.2

0.1

0.1

0.3

0.1

0.2

0.1

0.2

Case 2 0.3

0.1

0.1

0.1

0.2

0.1

0.2

0.1

0.2

As an example of the network structure under consideration (Fig. 2), we have the following set of paths between the first and seventh routers: • • • •

L1 L2 L3 L4

= {E1,2 , E2,5 , E5,7 }; = {E1,3 , E3,5 , E5,7 }; = {E1,3 , E3,6 , E6,7 }; = {E1,4 , E4,6 , E6,7 }.

Then, in the determination Mk , three variants of the solution (Table 2) for calculating disjoint paths can be obtained, the use of which provides the appropriate values of probability of compromising the multipath (8). Table 2. Results of the modelling. Multipath L 1 and L 3 L 1 and L 4 L 2 and L 4 Probability of compromising the multipath Case 1

0.1836

0.1836

0.2103

Case 2

0.1524

0.1836

0.1492

The application of the proposed model (1)–(8) made it possible to calculate the optimal multipath as the set of disjoint paths for each of the options for input data on the probabilities of network link compromising (Table 1). In the first case, the optimal solution is to use the paths L1 and L4 (Fig. 3) with the provision of PMP = 0.1836, the value of which is the minimum among the possible solutions (Table 2).

196

O. Lemeshko et al. R2 R5

R3

R1

R7

R6 0.1 R4

Fig. 3. Multipath for input data of the Case 1.

For the second variant of the input data (Table 1), the optimal solution is to use L2 and L4 paths (Fig. 4) with PMP = 0.1492, which is also the minimum of the three possible solutions (Table 2). R2 R5

R1

R7

R3 0.1 R6 0.1 R4

Fig. 4. Multipath for input data of the Case 2.

Thus, the solution of the relevant scientific and practical task related to the development of a mathematical model (1)–(8) for calculating the maximum number of disjoint paths during the implementation of secure routing in infocommunication networks was reduced to solving the ILP optimization problem with criterion (5) in the presence of linear constraints (1)–(4), since the routing variables are Boolean, and variables that determine the number of routes used take only integer values. The use of the metric (6) in the criterion of optimality (5) allows obtaining the maximum number of paths aimed at minimizing the probability of compromising the multipath. Thus, the application of the proposed model for calculating the maximum number of disjoint paths during the implementation of secure routing can increase the level of network security by the parameter of the probability of compromising the multipath.

System of Solutions the Maximum Number of Disjoint Paths Computation

197

4 Calculation Model the Set of Disjoint Paths with Maximum Bandwidth The basic model (1)–(4) can also be used in formulating the QoS routing problem. To do this, additional conditions will be introduced into its structure to ensure a given level of Quality of Service in terms of bandwidth. Let the value β be the bandwidth of the lowest-performing path included in the set of calculated disjoint routes, which can take real values. Then the following condition should be met (in analogy to [20]): k k ϕi,j + W (1 − ai,j ) ≥ β, ai,j

(9)

where the weighting coefficient W takes the value, which is higher than the maximum bandwidth of links Ei,j ∈ E belonging to the set of calculated disjoint paths available for the k th flow. The following function J2 , which should be maximized, has been chosen as the optimality criterion of the solutions to the problem of calculating the maximum number of disjoint paths in multipath routing under maximum bandwidth [24]:  k J2 = cM Mk + cβ β − cv vi,j ai,j , (10) Ei,j ∈E

where the weighting coefficients cM , cβ , and cv determine the importance of each of the components in the expression (10). The introduction of the first term in the criterion (10) maximizes the number of disjoint paths used. The second term is responsible for maximizing the bandwidth lower bound of the calculated disjoint paths. If we restrict ourselves to using only these two terms, then, on the one hand, the lowest-performing path will have a bandwidth equal to the β value. However, on the other hand, the use of only the first two terms in expression (10) does not always contribute to the inclusion in the calculated set of disjoint routes of paths that have the highest bandwidth. Therefore, the novelty of the proposed model is the use of the third term in the criterion (10), which is introduced by analogy with the metrics of the OSPF and EIGRP protocols in order to include in the calculated disjoint paths the links with high bandwidth. Therefore, it is proposed that in the objective function (10), the weighting coefficients vi,j under the bandwidth ϕi,j of the corresponding link Ei,j ∈ E are determined in the following way:  vi,j = 10 ϕi,j (11) Therefore, the task of solving the problem of calculation the set of disjoint paths with maximum bandwidth was reduced the to solving the optimization task of Mixed Integer Linear Programming (MILP) with criterion (10) in the presence of linear constraints and conditions (1)–(4), (9), (11) since the routing variables are Boolean, and variables of number of disjoint paths used take only integer values. In general, a variable β can be a real number, since bandwidth ϕi,j is not always expressed as an integer. It was established experimentally that the weighting coefficients in expression (10) must obey the following condition: cM  cβ  cv .

198

O. Lemeshko et al.

The features of the proposed improvement of the calculation model the set of disjoint paths with maximum bandwidth during multipath routing will be demonstrated on the following example. The structure of the analyzed network, which is shown in Fig. 5, consists of seven routers and eleven links. Let us consider that the first and seventh routers will be the corresponding source and destination nodes. In the example, it is supposed three cases of the values of links capacities in forming a set of disjoint paths when applying the proposed model for the initial data shown in Table 3. R2

R5 E2,5

R1

R7

R3 E1,3

E1,4

R4

R6 E4,6

Fig. 5. Network model.

On the network structure, which is taken into account in the investigation (Fig. 5), the following set of possible paths between the first and seventh routers has been obtained: • • • • • •

L1 L2 L3 L4 L5 L6

= {E1,2 , E2,5 , E5,7 }; = {E1,3 , E3,6 , E6,7 }; = {E1,4 , E4,7 }; = {E1,4 , E4,6 , E6,7 }; = {E1,3 , E3,5 , E5,7 }; = {E1,2 , E2,7 }.

Table 3. Initial data for investigation the calculation model of the set of disjoint paths with maximum bandwidth. Link Link bandwidth, 1/s Case 1 Case 2 Case 3 Case 4 E 1,2 300

220

800

200

E 1,3 900

210

100

300

E 1,4 150

240

150

150 (continued)

System of Solutions the Maximum Number of Disjoint Paths Computation

199

Table 3. (continued) Link Link bandwidth, 1/s Case 1 Case 2 Case 3 Case 4 E 2,5 250

190

250

250

E 2,7 400

200

400

100

E 3,5

90

180

890

270

E 3,6 110

190

310

110

E 4,7 140

210

140

190

E 4,6 300

185

300

300

E 5,7 920

180

220

220

E 6,7 180

190

780

180

The numerical research of the proposed model was conducted, the results of which for the initial data from Table 3 are presented in Table 4. In addition, the calculation of sets of disjoint routes was carried out under the condition of two options for using the optimality criterion (10): in the absence of the third term (Variant I) and taking into account all three components (Variant II). Table 4. Results of calculation the sets of disjoint paths between the first and seventh routers and path bandwidth. Case#

Set of disjoin paths

Bandwidth, 1/s

Links

Links

Links

I

{E 1,2 , E 2,5 , E 5,7 }

{E 1,3 , E 3,6 , E 6,7 }

{E 1,4 , E 4,7 }

500

II

{E1,3 , E3,6 , E6,7 }

{E1,4 , E4,7 }

{E1,2 , E2,7 }

550

Case 2

I

{E 1,3 , E 3,6 , E 6,7 }

{E 1,4 , E 4,7 }

{E 1,2 , E 2,7 }

600

II

{E 1,3 , E 3,6 , E 6,7 }

{E 1,4 , E 4,7 }

{E 1,2 , E 2,7 }

600

Case 3

I

{E 1,4 , E 4,7 }

{E 1,3 , E 3,5 , E 5,7 }

{E 1,2 , E 2,7 }

640

II

{E1,4 , E4,6 , E6,7 }

{E1,3 , E3,5 , E5,7 }

{E1,2 , E2,7 }

650

I

{E 1,2 , E 2,5 , E 5,7 }

{E 1,3 , E 3,6 , E 6,7 }

{E 1,4 , E 4,7 }

460

II

{E 1,2 , E 2,5 , E 5,7 }

{E 1,3 , E 3,6 , E 6,7 }

{E 1,4 , E 4,7 }

460

Case 1

Case 4

The calculations demonstrated that in two cases (Case 2 and Case 4) the introduction of the third term in criterion (10) did not affect the nature of the resulting routing solutions. Whereas in Case 1 and Case 3 (Table 4), the set of paths calculated by the integral criterion (10) had a higher bandwidth. For clarity, we will consider Case 3 in more detail (Fig. 6 and Fig. 7), for which we will show sets of disjoint paths for the corresponding initial (Table 3) and resulting (Table 4) data.

200

O. Lemeshko et al. R2

R5 250

R1

R7

R3 100

150 R4

R6 300

Fig. 6. Set of disjoint paths under criterion (10) without last term (Case 3).

Figure 6 shows the multipath calculated under the condition that the third term in the optimality criterion (10) is zero. In this case, the bandwidth of multipath is 640 1/s (Fig. 6). While the use of criterion (10) in calculations, taking into account all the terms included in its composition, makes it possible to obtain a set of disjoint paths with the maximum possible bandwidth of 650 1/s. R2

R5 250

R1

R7

R3 100

150 R4

R6 300

Fig. 7. Set of disjoint paths under criterion (10) (Case 3).

The analysis of calculation results given in Table 4 showed that the use of the integral criterion (10) allows providing higher bandwidth for the routing solution, represented by the set of disjoint paths, in cases of high heterogeneity of the network, i.e. when the bandwidths of the ICN communication links are quite different. This is typical for Case 1 and Case 3 (Table 4). With a homogeneous network architecture, when bandwidth of its links did not differ so much (Case 2 and Case 4 in Table 4), the introduction of the third term in criteria (10) did not affect the total bandwidth of the calculated disjoint paths, which determines the preferred scope of the proposed advanced routing solution.

System of Solutions the Maximum Number of Disjoint Paths Computation

201

5 Model of the Disjoint Paths Set Calculation for Secure QoS Routing Next, we use the basic model (1)–(4) to formulate the complex Secure QoS Routing task. In this case, conditions (9) will be used that provide a given level of Quality of Service in terms of bandwidth, as well as conditions for increasing the level of network security in terms of the probability of compromise the multipath, which in turn composed by disjoint paths (6)–(8). The following function J3 , which should be maximized, has been chosen as the optimality criterion of the solutions to the problem of calculating the maximum number of disjoint paths in multipath routing under maximum bandwidth in conjunction with the minimum value of the probability of compromise the entire multipath:  k wi,j ai,j (12) J3 = cM Mk + cβ β − cw Ei,j ∈E

where the weighting coefficients wi,j are determined by the expression (6). The novelty of the proposed model is the use of the third term in the criterion (12), which is introduced in order to include in the calculated disjoint paths the links with low probability of compromise. The features of the enhanced solution of the disjoint paths set calculation for Secure QoS Routing in the infocommunication network is presented on the following numerical example. At the structure of the analyzed network, which is shown in Fig. 5, the first and seventh routers will be the corresponding source and destination nodes. It is supposed three cases of the values of links probabilities of compromising in forming a set of disjoint paths when applying the proposed model (Table 5). Table 5. Initial data for investigation. Link

Bandwidth, 1/s

Case 1

Case 2

Case 3

Probability of compromise

Probability of compromise

Probability of compromise

E 1,2

800

0.4

0.2

0.1

E 1,3

100

0.3

0.1

0.1

E 1,4

150

0.4

0.3

0.1

E 2,5

250

0.2

0.2

0.1

E 2,7

400

0.2

0.4

0.2

E 3,5

890

0.1

0.1

0.2

E 3,6

310

0.2

0.1

0.2

E 4,7

140

0.2

0.1

0.2

E 4,6

300

0.2

0.3

0.1 (continued)

202

O. Lemeshko et al. Table 5. (continued)

Link

Bandwidth, 1/s

Case 1

Case 2

Case 3

Probability of compromise

Probability of compromise

Probability of compromise

E 5,7

220

0.3

0.2

0.1

E 6,7

180

0.1

0.4

0.1

The set of possible paths between the source and destination nodes will be used as in the previous example (Fig. 5). The numerical research of the proposed model was conducted, the results of which are presented in Table 6. Table 6. Results of calculation the sets of disjoint paths. Case#

Set of disjoin paths

Bandwidth

Probability of compromise

{E 1,2 , E 2,7 }

640

0.1341

{E 1,3 , E 3,5 , E 5,7 }

{E 1,2 , E 2,7 }

640

0.0677

{E 1,3 , E 3,5 , E 5,7 }

{E 1,2 , E 2,7 }

650

0.0267

Links

Links

Links

Case 1

{E 1,3 , E 3,6 , E 6,7 }

{E 1,4 , E 4,7 }

Case 2

{E 1,4 , E 4,7 }

Case 3

{E 1,4 , E 4,6 , E 6,7 }

The calculations demonstrated that in three cases the introduction of the third term in criterion (12) allowed to obtain the multipath with high bandwidth and minimal value of the probability of compromise the links comprised the multipath. For clarity, we will consider Case 1 in more detail (Fig. 5), for which we will show sets of disjoint paths for the corresponding initial (Table 5) and resulting (Table 6) data (Fig. 8).

Fig. 8. Set of disjoint paths (Case 1).

The analysis of calculation results given in Table 6 showed that the use of the integral criterion (12) allows providing higher bandwidth for the routing solution towards

System of Solutions the Maximum Number of Disjoint Paths Computation

203

maximization of the capacity of the lowest-performing path and minimum value of the multipath probability of compromise.

6 Conclusion In the course of the analysis, it was found that routing solutions based on the use of disjoint paths contribute to improving the level of Quality of Service, network security, and resilience in ICN. To obtain and use technological solutions represented by the relevant protocols and mechanisms, they should be based on adequate mathematical models and methods, consistent by ICN parameters for QoS, as well as network security and resilience. Therefore, due to the relevance of such a scientific and applied task, the paper proposes a system of solutions to the maximum number of disjoint paths computation under the Quality of Service and security parameters. The field of application of the calculation models the set of disjoint paths is in the provision of the network capabilities such as QoS and Network Security by the means of the OSI Network Layer, namely routing and traffic management protocols and technological solutions. Here secure routing means aimed at improving network security in terms of the probability of compromising, for example, the transmitted confidential data. Towards improving network performance and providing the demanded level of QoS in the network, the QoS-based multipath routing with support of load balancing and Traffic Engineering concept is used. The basic mathematical model for calculating the maximum number of disjoint paths has been presented by corresponding constraints and conditions (1)–(4). If necessary, the present model (1)–(4) can be used to calculate a given number of k ) for a particular packet flow or confidential data, which is achieved disjoint paths (Mreq k . The main computational advantage of the by introducing the condition Mk = Mreq basic model is the linearity of its constituent conditions, which greatly simplifies its subsequent protocol implementation. Depending on the particularities of the formulation of the disjoint multipath routing problem, the proposed model was supplemented by the optimality criterion for the obtained routing solutions and, if need be, additional constraint conditions. As a rule, the tasks stated were reduced to solving the optimization problems of integer and mixed integer linear programming. Therefore, when solving the secure routing problem to minimize the probability of compromising the calculated paths (multipath), a mathematical model for calculating the maximum number of disjoint paths (1)–(4) with the optimality criterion (5) and conditions (6)–(8) has been proposed. This task was reduced to the optimization of the ILP type. In addition, the choice of the weighting coefficients wi,j is based on the use of expression (6) and focused on the inclusion in a set of disjoint paths of links with a minimum probability of compromise. To solve the multipath QoS routing problem, a calculation model the set of disjoint paths with maximum bandwidth was developed, which was a further development of the basic model (1)–(4). The model of multipath QoS routing over disjoint routes included conditions (1)–(4), (9) and (11), as well as the optimality criterion (10). On the grounds that, the routing problem was reduced to the optimization of the MILP type. The analysis of calculation results showed that the use of the integral criterion (10) allows providing

204

O. Lemeshko et al.

higher bandwidth for the routing solution, represented by the set of disjoint paths, in cases of high heterogeneity of the network, i.e. when the bandwidths of the ICN communication links are quite different. Finally, a complex model of the disjoint paths set calculation for Secure QoS Routing was proposed, within the framework of which conditions (1)–(4), (6)–(9) were used together with the integral optimality criterion (12). Thus, this technological task was reduced to the optimization of the MILP type with maximization of the number of paths and their bandwidth and minimization of their compromising probability of entire multipath in the presence of linear constraints since the routing variables are Boolean, and variables that determine the number of routes used take only integer values. A number of numerical examples demonstrate the adequacy of the proposed system of solutions in terms of the correctness of the obtained calculation results. The proposed models can be used as an algorithmic basis for promising solutions regarding Secure QoS Routing. Moreover, their linearity is a significant advantage and aims to reduce the computational complexity of the final protocol solution.

References 1. White, R., Banks, E.: Computer Networking Problems and Solutions: An Innovative Approach to Building Resilient, Modern Networks. Addison-Wesley Professional, Boston (2017) 2. Barreiros, M., Lundqvist, P.: QOS-Enabled Networks: Tools and Foundations. 2nd edn. Wiley, Hoboken (2016) 3. Monge, A.S., Szarkowicz, K.G.: MPLS in the SDN Era: Interoperable Scenarios to Make Networks Scale to New Services. O’Reilly Media Inc., Sebastopol (2015) 4. Cisco Networking Academy (Ed.) Routing Protocols Companion Guide. Pearson Education (2014) 5. Yeremenko, O., Lebedenko, T., Vavenko, T., Semenyaka, M.: Investigation of queue utilization on network routers by the use of dynamic models. In: 2015 Second International Scientific-Practical Conference Problems of Infocommunications Science and Technology (PIC S&T) Proceedings, pp. 46–49. IEEE (2015). https://doi.org/10.1109/INFOCOMMST. 2015.7357265 6. Lemeshko, A.V., Evseeva, O.Y., Garkusha, S.V.: Research on tensor model of multipath routing in telecommunication network with support of service quality by greate number of indices. Telecommun. Radio Eng. 73(15), 1339–1360 (2014). https://doi.org/10.1615/Teleco mRadEng.v73.i15.30 7. Lemeshko, O.V., Yeremenko, O.S.: Dynamics analysis of multipath QoS-routing tensor model with support of different flows classes. In: 2016 International Conference on Smart Systems and Technologies (SST) Proceedings, pp. 225–230. IEEE (2016). https://doi.org/10.1109/ SST.2016.7765664 8. Guck, J.W., Van Bemten, A., Reisslein, M., Kellerer, W.: Unicast QoS routing algorithms for SDN: a comprehensive survey and performance evaluation. IEEE Commun. Surv. Tutorials 20(1), 388–415 (2018). https://doi.org/10.1109/COMST.2017.2749760 9. Koryachko, V.P., Perepelkin, D.A., Byshov, V.S.: Enhanced dynamic load balancing algorithm in computer networks with quality of services. Automat. Control Comput. Sci. 52(4), 268–282 (2018). https://doi.org/10.3103/S0146411618040077 10. Mohan, P.M., Gurusamy, M., Lim, T.J.: Dynamic attack-resilient routing in software defined networks. IEEE Trans. Netw. Serv. Manage. 15(3), 1146–1160 (2018). https://doi.org/10. 1109/TNSM.2018.2846294

System of Solutions the Maximum Number of Disjoint Paths Computation

205

11. Francois, F., Gelenbe, E.: Optimizing secure SDN-enabled inter-data centre overlay networks through cognitive routing. In: 2016 IEEE 24th International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (MASCOTS) Proceedings, pp. 283–288. IEEE (2016). https://doi.org/10.1109/MASCOTS.2016.26 12. Wang, M., Liu, J., Mao, J., Cheng, H., Chen, J., Qi, C.: RouteGuardian: constructing secure routing paths in software-defined networking. Tsinghua Sci. Technol. 22(4), 400–412 (2017) 13. Lou, W., Liu, W., Zhang, Y., Fang, Y.: SPREAD: improving network security by multipath routing in mobile ad hoc networks. Wirel. Netw. 15(3), 279–294 (2009) 14. Alouneh, S., Agarwal, A., En-Nouaary, A.: A novel path protection scheme for MPLS networks using multi-path routing. Comput. Netw. 53(9), 1530–1545 (2009) 15. Challal, Y., Ouadjaout, A., Lasla, N., Bagaa, M., Hadjidj, A.: Secure and efficient disjoint multipath construction for fault tolerant routing in wireless sensor networks. J. Netw. Comput. Appl. 34(4), 1380–1397 (2011) 16. Guo, L.: Efficient approximation algorithms for computing k disjoint constrained shortest paths. J. Combinatorial Optimizat. 32(4), 1–15 (2015). https://doi.org/10.1007/s10878-0159934-2 17. Yeremenko, O.: Enhanced flow-based model of multipath routing with overlapping by nodes paths. In: 2015 Second International Scientific-Practical Conference Problems of Infocommunications Science and Technology (PIC S&T) Proceedings, pp. 42–45. IEEE (2015). https:// doi.org/10.1109/INFOCOMMST.2015.7357264 18. Yeremenko, O., Lemeshko, O., Persikov, A.: Secure routing in reliable networks: proactive and reactive approach. In: Shakhovska, N., Stepashko, V. (eds.) CSIT 2017. AISC, vol. 689, pp. 631–655. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-70581-1_44 19. Lemeshko, O., Yeremenko, O., Persikov, A., Vavenko, T.: mathematical model of calculating the maximum number of disjoint paths in secure routing. In: 2018 International Conference of Information and Telecommunication Technologies and Radio Electronics (UkrMiCo) Proceedings, pp. 1–4. IEEE (2018) 20. Cruz, P., Gomes, T., Medhi, D.: A heuristic for widest edge-disjoint path pair lexicographic optimization. In: 2014 6th International Workshop on Reliable Networks Design and Modeling (RNDM) Proceedings, pp. 9–15. IEEE (2014). https://doi.org/10.1109/RNDM.2014.701 4925 21. Rak, J. (ed.): Resilient Routing in Communication Networks. CCN. Springer, Cham (2015). https://doi.org/10.1007/978-3-319-22333-9 22. Papan, J., Segec, P., Paluch, P., Uramova, J., Moravcik, M.: The new Multicast Repair (MREP) IP fast reroute mechanism. Concurrency and Computation: Practice and Experience, p. e5105 (2018). https://doi.org/10.1002/cpe.5105 23. Lemeshko, O., Yeremenko, O., Tariki, N.: Solution for the default gateway protection within fault-tolerant routing in an IP network. Int. J. Electric. Comput. Eng. Syst. 8(1), 19–26 (2017). https://doi.org/10.32985/ijeces.8.1.3 24. Lemeshko, O., Yeremenko, O., Yevdokymenko, M., Sleiman, B.: Improvement of the calculation model the set of disjoint paths with maximum bandwidth. In: 2019 International Conference on Information and Telecommunication Technologies and Radio Electronics (UkrMiCo) Proceedings, pp. 1–4. IEEE (2019)

Conditionally Infinite Telecommunication Resource for Subscribers Larysa Globa1(B)

, Mariia Skulysh1(B)

, and Eduard Siemens2(B)

1 National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic

Institute”, Kyiv, Ukraine [email protected], [email protected] 2 Anhalt University of Applied Sciences, Anhalt, Germany [email protected]

Abstract. Modern communication systems should provide the end user with a conditionally infinite telecommunication resource. To ensure guaranteed quality, services are combined into groups. A group of similar services with the same service process requirements is called a slice. Using software-defined networks (SDN) allows to deploy a management system slices and telecommunications resources that are allocated for their maintenance. This is possible only using the latest telecommunication technologies, and seamless user connection management. The article proposes an original methodology for managing the servicing streams process in network service providers tunnels. The methodology uses the allowable service node load calculation, overload prediction and live migration algorithms to seamlessly change the service node for the flow. Keywords: SDN · Micro operator network · 5G · Network slicing · Communication networks · Smart migration system

1 Introduction Nowadays, the number of Internet users are rapidly increasing along with the number of devices we use to get access to the Internet. The growing number of network services, more strict requirements to the quality of service and to the speed of the fixed and mobile Internet-access have caused the modernization of the network architecture, based on new requirements, standards and recommendations. Due to the active development of the Internet of Things concept, devices connected to the Internet generate huge volumes of network traffic, which makes the further usage of the networks built only on routers and switches impossible. It becomes one of the most important reasons of active development of the Software Defined Networks (SDN) and Network Functions Virtualization Services (NFVS), which are integral elements of the fifth generation network concept [1]. With these technologies, 5G networks can provide the necessary programmability, flexibility, scalability, etc., that are needed to build different logical (virtual) networks for a specific type of task, without changing the infrastructure platform. These logical © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 206–216, 2021. https://doi.org/10.1007/978-3-030-58359-0_11

Conditionally Infinite Telecommunication Resource for Subscribers

207

networks with shared resources are network slices. The purpose of this work is to analyze the current models and methods of network slicing in modern telecommunication networks, and their implementations. A significant part of the new operators prefers to cooperate with small microtelecommunication networks that cover the interior after the onset of the 5G era. Since these operators can provide various networks such as 3G, 4G, 5G and even Wi-Fi [1, 2]. Up-to-day technology use more and more Network Function [7] that is a functional unit within a network infrastructure that has clearly defined external interfaces and welldefined functional behaviour. In practice, a network function is today a network node or physical device. Innovative industry groups such as the ETSI ISG group (Industry Specification Group) for the NFV and ONF (the Open Networking Foundation) organization for the SDN created reference architectures, substantiated usage scenarios and changed the requirements for the components that are an integral part of NFV and SDN. The SDN network architecture supports the principles of network slicing because SDN allows you to manage a common infrastructure network and effectively support multiple client instances of the network [8]. Figure 1 illustrates an example of integrating SDN and NFV technologies for implementing slicing in 5G communication networks [9].

Fig. 1. Network slice architecture for SDN and NFV

208

L. Globa et al.

2 State of the Art and Background The task is to organize the maintenance of independent slices within the existing computation and telecommunication infrastructure. To ensure the maintenance of independent slices, the following features must be considered: • it is required to provide each with sufficient amount of resources, both telecommunication and computing ones, for servicing virtualized network functions in order to provide service at a given quality level. • it is necessary to take into account the nature of the load change in each slice for optimal allocation of resources between slices during the day. • determine the conditions for the migration of slices in the telecommunications infrastructure, which will ensure the smooth operation of the system. With regard to infrastructure design, there SDN/NFV network technology and tunnels to slice services are used. Infrastructure allows users to connect multiple interfaces using tunneling technology and running fast network connection to effectively strengthen the relationship networks. In response to the demand telecom operator network resource distribution that allows users to gain access of nearby network resources, the paper [8] proposes network selection mechanism for a Micro Operator and uses decision tree theory to serve as the reference in determining the SDN traffic flows path. The proposed method disadvantage is that traffic will be distributed without taking into account the all network resources load. The method shown in Fig. 2 application will allow to predict the moments network tunnels overload and to migrate the slice’s sessions between network tunnels in time, which will allow to provide conditionally infinite bandwidth for each network slice. The “The flows and node resources use monitoring” block involves the accumulation the communication channels congestion information, and information about the resource use dynamics by each service. For the ediction of the possibility of exceeding the permissible value λ (the admissible input stream intensity), the method proposed in [15] is used. The basic method idea is to formulate requirements for the average input load on the basis of ergodic distribution for the possible states of the system, which will allow to make the most efficient use of the available physical resources of servicing the incoming application flow. For prediction of exceeding the permissible value λ we propose to use the method [7] consists of two stages: the calculation of the prediction interval based on the operation servicing node statistics and directly periodic forecasting of the load and the control of the sufficiency resources. If periodic forecasting of the load showed that it overload is expected and the available slice resources are not enough to provide services at a given level, then the migration mechanism starts.

Conditionally Infinite Telecommunication Resource for Subscribers

209

The optimal node load (λmax) calculation The flows and node

Prediction of

resources use

exceeding the

monitoring

permissible

Flow migra-

Flow migra-

tion

tion capability estimation

Fig. 2. Dynamic flow control algorithm

3 The Method of Forming Input Load Flow for Efficient Use of Service Resources The main method objective is to formulate requirements for the average input load on the basis of ergodic distribution for the possible system states, which will allow to make the most efficient available physical resources of servicing the incoming flow. The servicing process is modeled as an n-channel servicing device, the service time in the channel is a random variable distributed by Poisson law. A. Input data n is number of channels for simultaneous requests service. μ is the service intensity, G – is the number of resources involved in servicing slices, vg – the amount of g-resource required to serve in a single request, g = 1, G. V g – d is the available volume of the g-th resource shared by requests. s is the allowable number of service queue requests. R is the percentage of applications served on the system not exceeding the allowable delay time. l is the number of queued requests that block requests from being sent to the system, according to early overload prevention algorithms. B. Problem It’s necessary to find the recommended value for the input stream intensity (λ). Application of the proposed method consists of two steps.

210

L. Globa et al.

Step 1. According to K. Zhernovyi’s models, find ergodic distribution of the number of applications in the system for multi-channel service system, by the formulas: p0 =

1 − β s−l , An (α, β)

β = 1,

α = λ/μ, β = λ/n

  n   α n β − β s−l+1 αk s−l s+1 + − (s − l)β An (α, β) = 1 − β k! n! 1−β k=0

pk =

αk p0 k!

pn+k =

  k = 1, n

  αn k β p0 k = 1, l n!

 αn β k − β s  k = p l + 1, s − 1 0 n! 1 − β s−l  α n (1 − β)β s  k = n + l + 1, n + s − 1 = n! An (α, β)

pn+k = pn+s

where β = 1, α = λ/μ, β = λ/n, and     k n pk = nk! p0 k = 1, n ; pn+k = nk! p0 k = 1, l ; pn+s =

nn n! p0 (s − l

− 1)pn+s ⇒

⇒ pn+s = pn+k =

  nn p0 k = n + l + 1, n + s − 1 ; n!(s − l)

  nk nn s − k p0 − (k − l)pn+s = p0 k = l + 1, s − 1 n! n! s − l

Step 2. Solving the optimization problem of finding the maximum load, which will ensure the fulfilment of conditions for an acceptable amount of service resources. λ → max  ⎧  n s ⎪

g g ⎪ ⎪ ivk pi + nvk pi ≤ V g , g = 1, G ⎨4 ∗ ⎪ ⎪ ⎪ ⎩

i=1

i=n+1

s

i=1

pi ≤ R

Incoming stream balancing in a billing system is performed using a modified Round Robin scheme. To calculate the recommended input intensity value of the entire system, it is sufficient to analyze the metrics from one DOCS server and scale the obtained values to the entire subsystem.

Conditionally Infinite Telecommunication Resource for Subscribers

211

Calculation of the Optimum Value of the Input Stream Intensity for an Existing System To calculate the optimal value of λ, we analyzed the metrics available in the billing subsystem for the busiest day in 2017 (11/24/2017 Black Friday). The maximum input stream intensity for a known number of sessions was calculated at peak hours, at the time when the service failure metrics (REJECTS) were signaled. The analysis the service intensity of the application (μ) and the concurrent value of the input stream intensity (λ) for each record was calculated. Based on the data obtained, we can conclude that the maximum allowable value of the input stream intensity for a single OCS process in which no service degradation occurs (no REJECTS) corresponds to λ = 860 at μ = 430. A graphical representation of the calculated application service intensity (μ) and the simultaneous input stream intensity value (λ) for a single OCS process is shown in Fig. 3.

Fig. 3. Graph of the intensity value of λ for 11/24/2017

Short-Term Load Forecasting Method Short-term scheduling is an advanced ARIMA prediction method - an autoregressive method with a rolling expectation. However, unlike the known method, it is suggested to solve the problem of finding a minimum rolling interval, the use of which will satisfy the requirements. This will minimize the number of floating point operations to perform the prediction, which will provide the optimum speed of the prediction execution. The proposed method consists of two steps - the calculation of the forecasting interval based on the statistics of operation of this service node and directly periodic load forecasting and control of resource adequacy. Incoming Data: • Tp - the time interval for which the forecast is required. • λi - number of applications per 1 ms, (i  0, .. N), N = T inf /1 ms, λi  , where  - the set of statistics for the number of applications received over time Tinf (it is first specified, then corrected in step 2 of the method) before the forecast, || = N.

212

L. Globa et al.

• Tfor - forecasting period, the time after which, runs forecasting algorithm. • M - the maximum number of requests that can be served with a given configuration of the serving device. • P - the probability of a prediction error. Short-term statistics is collected locally with the operating device and stored no longer than Tinf + Tfor , a sampling interval of 1 ms. Output Data: • z  {0,1} – z = 0 do not change the configuration; z = 1 change the configuration. Method Algorithm: Preparatory Stage. Learning the system based on statistics. Finding the minimum Tinf (Information Collection Time Interval) Tinf → min for which the limit is performed: λ¯ Tfor + 3σ > M , where λ¯ Tfor and σ – are calculated according to the main step. The restriction is satisfied for P statistical samples obtained at different time intervals. k Solution method: validate values for a sequence formed by a principle Tk+1 nf = Tinf + ; T0inf = Tfor . The Main Stage of Dynamic Control. 1. Analysis of statistics λi, for the time interval Tinf , preceding the moment of calculation. Construction of coefficients aˆ and bˆ by the least square method: λ = ai + b ˆ 2. Calculate λ¯ Tfor = aˆ Tfor + b. 3. If λ¯ Tfor + 3σ ≤ M , than z = 0, else z = 1.

4 Method of Automatic Live Migration The main task is to provide automatic balancing of the load of physical resources in telecommunication nodes. In this case, it is necessary to avoid too much load on one node and inefficient resources use of the second one. This balancing mechanism is called “smart migration”.

Conditionally Infinite Telecommunication Resource for Subscribers

213

We formulate the main tasks of the system: 1. Make the migration process invisible to users. 2. Use the migration mechanism in order to optimize the state of the telecommunication network. 3. Ensure a minimum migration time, since too long process will negatively affect the state of the system. 4. The operative and backup channel of one slice should be located in different physical nodes. For this purpose, it is necessary to ensure high availability of the telecommunication node. 5. Provide anti-loop protection to prevent endless migration of the same slice. 6. Provide protection against failures. 7. Work in cluster mode in the case of multiple migrations.

Migration System Architecture Each slice forms data from the load statistics on the telecommunication node. They are periodically read and converted to metrics, which are placed in a specific storage. Having access to such a repository, one can analyze the dynamics of resource consumption in each telecommunication node. It is also possible to get information about the amount of physical servers’ resources. The control unit makes a decision about the need for migration and distributes the load between nodes (Fig. 4). The sequence of the process is as follows:

Fig. 4. Migration of flows between the neighboring micro-operator networks

1. First, candidates for migration are selected and placed in a distributed queue. 2. The queue is processed by a special process that analyzes the number of elements in it and selects the one to be transferred. Physical migration is a synchronous process without breaks and returns. A special mechanism allows to avoid delays and errors. The main task of the process is to choose the right candidate for migration and transfer it to the desired system node,

214

L. Globa et al.

ensuring its optimal functioning. For this purpose, it is necessary to solve the problem of multidimensional optimization. There are several algorithms for solving the problem. Simple Algorithm 1. Analyze the load on each node. 2. Select the node that uses resources most optimally. 3. Move the slice to this node. This algorithm will work effectively with a small number of telecommunication nodes. However, for a large network, it is not optimal. In this case, you will need a more complex algorithm with hard and soft restrictions. Its essence is to determine two kinds of rules: those that cannot be violated in any case, and those that can be neglected under certain circumstances. You also need to identify three types of problem solutions: 1. Possible solutions – bad decisions that violate hard rules. 2. Feasible solutions - violate some of the soft rules, but do not violate hard rules. 3. Optimal solutions - satisfy both hard and soft restrictions. These are the best decisions received as soon as possible. First, we define hard and soft restrictions for a telecommunication system that needs to be optimized: Hard Restrictions 1. The target node must have enough resources to switch the slice. You also need to provide redundancy resources. 2. It is not allowed to move a slice to its own physical node. 3. A physical node cannot contain flows of the same slice. Thus, in the event of a system failure, there will be minimal data loss.

Soft Restrictions 1. It is necessary to ensure the migration of the most loaded slice streams. 2. The target physical node must have minimal load. This algorithm has one significant disadvantage: it does not have tools that will take into account trends in the rate of consumption of resources by different slices. We conducted a study on the effectiveness of processing statistical data during the selection of the migration flow and the target node. For the correct choice of the node to which the flows will be redirected, it is necessary to evaluate the dynamics of the resources usage of a particular server. This takes into account the load created by the flow of migrating slices.

Conditionally Infinite Telecommunication Resource for Subscribers

215

To determine the moment of migration, it is necessary to evaluate the current statistics of resource use with a given time interval and build a statistical trend, taking into account the number of served requests. Based on this trend, the probability that the content of the containers of the evaluated node will exceed the available amount of resources is estimated. Then the migration process begins. A method for assessing the adequacy of resources is presented in [10]. Based on the statistics received, the sum of the flows of the selected slices will be generated. After evaluating the upper limit of the node capacity, it will be possible to make a decision on the need for migration.

5 Conclusion This document uses SDN and NFV technologies as the basis and combines network streaming and tunneling technologies to create a network infrastructure model for MSO using a smart migration mechanism. This model allows users of different MOs to connect using tunneling technology, and then implement a fast network connection to effectively improve network interaction, while balancing the load between all nodes of a given network. To meet the needs of the regional micro-operator service, this article proposes a DTBFR mechanism that uses decision tree theory as the basis for making SDN-based traffic decisions. As a distribution and control of the load on the nodes, we use the method of slices working together in the existing telecommunication foreign infrastructure, which ensures the automatic distribution of telecommunication and computing resources of the system depending on the load and allows solving the problem of peak loads and idle resources. This method of automatic load balancing of telecommunication nodes (“smart migration”) does not allow overloading one node and downtime of another node. The functions used by the regional micro-operator service can effectively reduce the load on the datacenter on the Internet and accelerate the development of the regional computer service in the future 5G network. And the “smart migration” method will allow rational use of network resources.

References 1. Raschellà, A., Bouhafs, F., Deepak, G.C., Mackay, M.: QoS aware radio access technology selection framework in heterogeneous networks using SDN. J. Commun. Netw. 19(6), 577– 586 (2017) 2. Matinmikko, M., Latva-aho, M., Ahokangas, P., Yrjölä, S., Koivumäki, T.: Micro operators to boost local service delivery in 5G. Wirel. Pers. Commun. 95(1), 69–82 (2017) 3. Walia, J.S., Hammainen, H., Matinmikko, M.: 5G micro-operators for the future campus: a techno-economic study. In: Proceedings of the 2017 Internet of Things - Business Models, Users, and Networks, Copenhagen, Denmark, pp. 1–8, November 2017 4. Matinmikko-Blue, M., Latva-aho, M.: Micro operators accelerating 5G deployment. In: Proceedings of the 2017 IEEE International Conference on Industrial and Information Systems (ICIIS), Peradeniya, Sri Lanka, pp. 1–5, December 2017 5. Ahokangas, P., Moqaddamerad, S., Matinmikko, M.: Future micro operators business models in 5G. Bus. Manag. Rev. 7(5), 143–149 (2016)

216

L. Globa et al.

6. Mach, P., Becvar, Z.: Mobile edge computing: a survey on architecture and computation offloading. IEEE Commun. Surv. Tutor. 19(3), 1628–1656 (2017) 7. Globa, L., Skulysh, M., Romanov, O., Nesterenko, M.: Quality control for mobile communication management services in hybrid environment. In: The International Conference on Information and Telecommunication Technologies and Radio Electronics, pp. 76–100. Springer, Cham, November 2018 8. Skulysh, M.A., Romanov, O.I., Globa, L.S., Husyeva, I.I.: Managing the process of servicing hybrid telecommunications services. Quality control and interaction procedure of service subsystems. In: International Multi-Conference on Advanced Computer Systems, pp. 244– 256. Springer, Cham, September 2018 9. Tseng, C.-W., Huang, Y.-K., Tseng, F.-H., Yang, Y.-T., Liu, C.-C., Chou, L.-D.: Micro operator design pattern in 5G SDN/NFV network. Wirel. Commun. Mob. Comput. 2018 14 (2018). Article ID 3471610 10. Semenova, O., Semenov, A., Voitsekhovska, O.: Neuro-fuzzy controller for handover operation in 5G heterogeneous networks. In: 2019 3rd International Conference on Advanced Information and Communications Technologies (AICT), pp. 382–386. IEEE, July 2019 11. Skulysh, M.: The method of resources involvement scheduling based on the long-term statistics ensuring quality and performance parameters. In: 2017 International Conference on Information and Telecommunication Technologies and Radio Electronics (UkrMiCo), pp. 1–4. IEEE, September 2017 12. Skulysh, M., Romanov, O.: The structure of a mobile provider network with network functions virtualization. In: 2018 14th International Conference on Advanced Trends in Radioelectronics, Telecommunications and Computer Engineering (TCSET), pp. 1032–1034. IEEE (2018) 13. Globa, L., Kurdecha, V., Ishchenko, I., Zakharchuk, A., Kunieva, N.: The intellectual IoTsystem for monitoring the base station quality of service. In: 2018 IEEE International Black Sea Conference on Communications and Networking (BlackSeaCom), pp. 1–5. IEEE, June 2018 14. Romanov, O.I., Hordashnyk, Y.S., Dong, T.T.: Method for calculating the energy loss of a light signal in a telecommunication Li-Fi system. In: 2017 International Conference on Information and Telecommunication Technologies and Radio Electronics (UkrMiCo), pp. 1–7. IEEE, September 2017 15. Skulysh, M., Shilov, F., Safaryan, A.: Investigation of the method of computing resources optimal choice for billing systems effectiveness. Control Navig. Commun. Syst. Acad. J. 3(49), 147–152 (2018). https://doi.org/10.26906/SUNZ.2018.3.147 16. Globa, L.S., Ishchenko, I., Kurdecha, V., Zakharchuk, A., Zvonarov, O.: An approach to the Internet of Things system with nomadic units developing. In: 2016 IEEE International Black Sea Conference on Communications and Networking (BlackSeaCom), pp. 1–5. IEEE, June 2016

Different Extrapolation Methods in Problems of Forecasting Irina Strelkovskaya(B)

, Irina Solovskaya(B)

, and Anastasiya Makoganiuk(B)

O.S. Popov Odessa National Academy of Telecommunications, Odesa, Ukraine {strelkovskaya,i.solovskaya,a.makoganyuk}@onat.edu.ua

Abstract. The task of forecasting characteristics of self-similar traffic in IoT network objects with a significant number of pulsations and the property of long-term dependence is considered, which makes it difficult to forecast in practice. Using the different extrapolation method of spline-extrapolation based on linear, cubic and B-cubic spline function and wavelet-extrapolation based on Haar-wavelet, the results of forecasting of self-similar traffic are obtained. The comparison made allowed the results of traffic forecasting based on the Haar-wavelet and the linear, cubic and B-cubic spline-function using wavelet- and spline-extrapolation. This will allow you to choose one or another extrapolation method to improve the accuracy of the forecast, while ensuring scalability and the ability to use it for various IoT applications to prevent network overloads. Keywords: Self-similar traffic · Internet of Things · Forecasting · Extrapolation · Spline-function · Wavelet-function

1 Introduction The development of IoT (Internet of things) networks based on Machine-toMachine Communication (M2M) is associated with the rapid introduction of a highspeed services nomenclature. These are Video Surviliance video services, SmartTV, IoT-cameras, smart-M2M objects services, IoT-telematra telemetry services, IoTtelemedicine telemedicine services, Smart house systems, smart parking, Smart parking lots, Smart grids and, in perspective, holographic TV, virtual VR (Virtual reality) and augmented reality (AR) services. Implementing IoT applications uses Wi-Fi radio access technologies (IEEE 802.11n/ac/ad standards), ZigBee and Bluetooth (IEEE 802.15 standards), as well as the LoRaWan (Low-power Wide-area Network) and NB-IoT mobile communication network (Narrow Band) fourth generation 4G/LTE and fifth generation NR (New Radio) technology 5G. A feature of the IoT network architecture is the use of AdHoc/Mash architectures, which are based on self-organization and the use of virtualization technologies, including cloud [1, 2]. The traffic that is generated by IoT network objects is quite heterogeneous, has different characteristics depending on the type of service. Sometimes this is a low-speed data transmission by IoT sensors, and sometimes, real-time HD-video transmission. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 217–228, 2021. https://doi.org/10.1007/978-3-030-58359-0_12

218

I. Strelkovskaya et al.

It is known [3, 4] that the high-speed data traffic that is generated by the IoT network objects is self-similar. This is determined by the long-term relationship between traffic values at different periods of time, significant and frequent bursts of intensity, which are statistically similar for different time scales. Thus, the IoT network in one and the same application may generate traffic using different M2M and D2D devices. Therefore, it is rather difficult to predict the impact of traffic of individual M2M and D2D devices on the network infrastructure of IoT. From the operator’s point of view, this does not allow for accurate analysis and evaluation of network performance and, accordingly, dynamic optimization of IoT network resources. In addition, according to [1], the IoT network architecture will be horizontal and will combine various open platforms and systems, but how this affects the traffic characteristics is not known yet. Solving the problem of forecasting self-similar IoT network traffic, both short-term and long-term, will allow you to get the traffic characteristics of various applications and M2M, D2D network devices. This will allow them to provide the necessary performance and the possibility of using for various IoT applications in order to prevent overloads and maintain the required quality of service QoE (Quality of Equipment). The issues of forecasting self-similar traffic in the IoT network are considered in the works by the authors [5–17]. The proposed solutions are not universal prediction, but allow solving many problems associated with forecasting. In [5], the authors considered the prediction of “IoT traffic paths” in real time and compared four prediction methods, such as the classical Fourier time series, artificial neural networks, and the double exponential smoothing method. This paper focuses not only on the accuracy of the comparison, but also on other factors: such as - the cost of solutions and energy consumption. On the contrary, in [6] IoT traffic prediction based on a recurrent neural network was proposed. In practice, it is rather difficult to forecast dynamic network update in real time using a neural network. An alternative to the considered methods is the “deep intelligent network study method” based on the prediction of traffic load channels of IoT devices [7]. In this case, the method allows avoiding overloads in the network and achieving optimization of network resources, but it requires specific “training” for each considered IoT network architecture. The results of traffic prediction using ARIMA/FARIMA models were obtained in [8]. However, these models do not allow for different types of traffic to obtain an accurate prediction of characteristics outside the considered interval. This is due to the fact that it is quite difficult to obtain the corresponding trend of a specific type of traffic M2M of IoT network device. Previously, the authors of the work obtained the results of short-term prediction of self-similar traffic using the spline-extrapolation method [9–13]. The use of cubic and cubic B-splines showed the advantage of using the latter, especially for real-time forecasting. The authors in [14–17] considered the use of wavelet functions for forecasting the characteristics of networks that are different in construction and operation. In each of the cases considered the use of the corresponding wavelets significantly improved the results obtained.

Different Extrapolation Methods in Problems of Forecasting

219

Therefore, in this paper, it is proposed to use the wavelet-extrapolation method to forecast the traffic of IoT network objects. This will improve the accuracy of the forecasting, ensure its scalability and use for various IoT applications. The goal of the work is to solve the problem of forecasting the characteristics of self-similar traffic of IoT network objects using wavelet extrapolation and to conduct a comparative analysis of this method with the previously considered spline-extrapolation method.

2 Prediction of the Self-similar Traffic Based on the Spline-Extrapolation To forecast the self-similar traffic, we use the extrapolation method. The research task is as follows [11]: 1) forecast the traffic beyond the considered time period, where the transmission of packet data is considered; 2) choose an approximating device, with the help of which it is possible to perform a more accurate prediction of self-similar traffic; 3) determine the error in restoring self-traffic outside the gap. Earlier, problem solving extrapolation of random processes was based on the use of Lagrange interpolation polynomials, Chebyshev polynomials, etc. As an approximating apparatus for forecasting the applicable “spline-extrapolation” when extrapolation is performed using spline-functions. The task of forecasting self-similar traffic is solved by the “spline-extrapolation” method, that is, the recovery of self-similar traffic outside the considered time interval based on spline functions (for example, linear, cubic, etc.). Considering that self-similar traffic is characterized by the presence of “bursts” packets, it is possible to use extrapolation based on wavelet-functions to improve the accuracy of prediction. Let us dwell on the method of “spline-extrapolation.” The term “splineextrapolation” we mean an extrapolation based on the use of spline functions. Consider self-similar traffic on the segment [a; b]. Let at the interval [a; b] is given a partition : a = x 0 < x 1 < … < x N = b. The first-degree spline S 1 (x) on the grid  - is a continuous piecewise linear function. Let  grid points are given the values of self-similar traffic f i = f (x i ), which describes a function f (x), defined on the interval [a; b]. The interpolation spline is defined by the following conditions: S1 (xi ) = fi , i = 0, . . . , N .

(1)

Geometrically, it is a broken line passing through the points (x i , yi ), where yi = f (x i ). We denote by hi = x i+1 − x i . Then, according to [18], for x ∈ [x i , x i+1 ], i = 0, …, N − 1, the linear spline will have the form: S1 (x) = fi

xi+1 − x x − xi + fi+1 . hi hi

(2)

220

I. Strelkovskaya et al.

or S1 (x) = fi +

x − xi (fi+1 − fi ). hi

(3)

We will also consider a cubic interpolation spline S 3 (x), constructed similarly to a linear spline, with the only difference being that this is a cubic function on each interval [x i , x i+1 ], i = 1, …, N − 1. According to [18], for x ∈ [x i , x i+1 ], i = 0, 1,…, N − 1 the cubic spline has the form: S3 (x) = fi (1 − t)2 (1 + 2t) + fi+1 t 2 (3 − 2t) + mi hi t(1 − t)2 − mi+1 hi t 2 (1 − t), (4) where t = (x − xi )/hi , S3 (xi ) = fi , S3 (xi+1 ) = fi+1 , mi = S  (f ; xi ) or S3 (x) = fi (1 − t) + fi+1 t −

h2i t(1 − t)[(2 − t)Mi + (1 + t)Mi+1 ] 6

(5)

where S  (xi ) = Mi , S  (xi+1 ) = Mi+1 . The boundary conditions are used to determine the cubic spline of the form (4) on the interval [a; b] [18]: S  (f ; a) = f  (a), S  (f ; b) = f  (b).

(6)

For determining a cubic spline type (5) using the boundary conditions of the form [18]: S  (f ; a) = f  (a), S  (f ; b) = f  (b).

(7)

We consider a uniform partition of the interval [a; b], that is: hi = h =

b−a , i = 0, 1, . . . N − 1. N

It is necessary to restore the self-similar traffic outside the interval [a; b], namely, the right of the point b. For definiteness, let this be a point x c , x c > x N = b, x c – b = h,

f

f(x0) f(x1)

h a=x0 x1 x2

b=xN xc

x

Fig. 1. The extrapolation of self-similar traffic on the interval [a; b] on condition f (xc ) = f (x1 ).

Different Extrapolation Methods in Problems of Forecasting

221

where h - is a step partition of the interval [a; b]. Construct on the interval [b; x c ] spline function (linear or cubic). Let’s consider two variants. In the first case, we assume that f (x c ) = f (x 1 ) and construct a linear or cubic spline, respectively, on the interval [b; x c ] (Fig. 1). In the second case (Fig. 2) we set f (x k ) = f (x c ), where x k = x 0 + kh, where k is natural. If kh = x c − b, then for f (x c ) we take the value of the function f (x) which are closest to the point x k .

f

f(xk) kh

f(x0) a=x0 x1 x2

b

xk

xc

x

Fig. 2. The extrapolation of self-similar traffic on the interval [a; b] on condition f (xkh ) = f (xc ).

According to [18], it is not difficult to find the error of restoring self-similar traffic on the interval [b; x c ], using the following theorems: Theorem 1 [18]. If a spline of the first degree S 1 (x) interpolates a continuous function f (x) on the grid , then the estimation of the error is valid: |S1 (x) − f (x)| ≤ ω(f ), where ω(f ) - module of a continuous function of the form   ω(f ) = max ωi (f ) = max  ,  max f (x ) − f (x ). 0≤i≤N −1

0≤i≤N −1x x ∈[xi ,xi +1]

(8)

(9)

Theorem 2 [18]. If the spline cubic S 3 (x) interpolates the continuous function f (x) on the net  and satisfies the boundary conditions (6) or (7), then 3 S (x) − f (x)C ≤ (1 + ρ)ω(f ), (10) 4 where ρ = max hi / min hi , f (x)C = max |f (x)|, C = C[a; b] - this process is i

i

x∈[a;b]

continuous on the interval [a; b] of the function.

3 Prediction of the Self-similar Traffic Based on the Wavelet-Extrapolation To predict self-similar traffic, we use the extrapolation method. The authors of [11–13] used an extrapolation method. When forecasting self-similar traffic, similarly to [11–13], we use the extrapolation method based on wavelet functions, since:

222

I. Strelkovskaya et al.

1) the use of wavelet functions allows to obtain information about the nature of changes in the parameters of the function under study, which is especially important when studying non-stationary traffic characteristics; 2) wavelet functions are able to identify local features of the studied self-similar traffic (break points, intensity bursts, sharp drops and self-similarity), which are often overlooked by other extrapolation methods, which allows to reproduce the most important characteristics of the traffic in question; 3) The use of wavelet functions allows independent analysis of the function at different scales of its change, because each traffic component is additionally described by a scaling function, which allows determining the point in time at which one or another traffic characteristic changed. The problem of forecasting self-similar traffic is solved by the wavelet-extrapolation method, that is, the recovery of self-similar traffic outside the considered time interval based on wavelet functions (for example, Haar-wavelet, Daubechies-wavelet, Morletwavelet, etc.) [19]. Let us dwell on the method of “wavelet-extrapolation.” By “wavelet-extrapolation” we mean extrapolation based on the use of wavelet functions. In work [20–23], wavelet transform allows to perform the signals restoration both in the temporary parameters with the basic wavelet function ψ(x), and in the frequency parameters with the scaling function ϕ(x): √  √  gk ϕ(2x − k), ϕ(x) = 2 hk ϕ(2x − k), (11) ψ(x) = 2 k

k

√  where hk = 2 ϕ(x)ϕ(2x − k)dx, gk = (−1)k h(N −1)−k−1 is expansion coefficient, k, N is natural numbers. With a special choice of the coefficients hk , we obtain mother wavelets of a particular type, which form an orthonormal basis. All the wavelets are obtained from the basic wavelet with the scaled transformation 1/2k and shifts j/2k [21, 22]: ψj,k = 2j/2 ψ(2j x − k), ϕj,k = 2j/2 ϕ(2j x − k).

(12)

Then, according to [20–23], any function f (x) from Hilbertian space L 2 (R) can be expressed in the series of the form:   sjn ,k ϕjn ,k (x) + dj,k ψj,k (x), (13) f (x) = k

j≥jn ,k

where f (x) is temporary function of the original signal, ϕ j,k (x) is scaling-function (father wavelet), ψj,k (x) is basic wavelet function (mother wavelet), sj,k and d j,k is wavelet coefficients, jn is scaling level, k, j, jn is natural numbers. Wavelet coefficients sj,k and d j,k for the expression (3) are calculated as follows [20–23]:   sj−1,k = hm sj,2k+m , dj−1,k = gm sj,2k+m , (14) m

m

Different Extrapolation Methods in Problems of Forecasting

223

where hm , gm is approximation and expansion coefficients. On the most detailed level of scaling jn = jmax there left only s coefficients and the signal is given by the scaling function [18–21]:  f (x) = sjmax ,k ϕjmax ,k (x). (15) k

Similar to [20], we consider the extrapolation of self-similar traffic using the simplest Haar-wavelet functions. It is known [14] that the Haar-wavelet is a short square wave on the interval [0, 1], which has a long ‘tail’ in the frequency domain and can be used for traffic. For self-similar traffic of IoT objects, we use Haar-wavelet, for which the basic wavelet function ψ(x) and scaling function ϕ(x) have the following form (Fig. 3,a) and (Fig. 3,b) respectively: Haar-wavelet

Haar scaling-function

1.5

1.5

0

0

− 1.5 −1

0.5

2

а)

− 1.5 −1

0.5

2

b)

Fig. 3. Mother Haar-wavelet ψ(x) a) Haar-wavelet, b) scaling function ϕ(x).

For a Haar wavelet, according to [20–23], the analytical record for its basic ψ(x) and scaling function ϕ(x) has the form: ⎧ ⎨ 1, 0 ≤ t < 1/2, ψ(t) = −1, 1/2 ≤ t < 1, (16) ⎩ 0, t < 0, t ≥ 1.  1, 0 ≤ t < 1, ϕ(t) = (17) 0, t < 0, t ≥ 1. The mother wavelet ψ(x) and the scaling function ϕ(x) refer to orthogonal subspaces of the hibert space L 2 (R), which corresponds to the fulfillment of the condition [20–23]:  hk gk+2M = 0. (18) k

From the formula (18) you can determine the dependence between the coefficients hk and gk of the Haar-wavelet values [20–23]: gk = (−1)k h2M −1−k .

(19)

224

I. Strelkovskaya et al.

The resulting hk and gk coefficients determine the Haar-wavelet. To perform the wavelet transform, the Haar calculation of the wavelet coefficients of the function in question is performed according to formulas (14), we obtain the formulas for the direct [20–23]



1 1 sj−1,k = √ sj,2k + sj,2k+1 , dj−1,k = √ sj,2k + sj,2k+1 , 2 2

(20)

and inverse transform



1 1 sj,2k = √ sj−1,k + dj−1,k , sj,2k+1 = √ sj−1,k − dj−1,k . 2 2

(21)

Thus, the calculated values of the coefficients of the Haar-wavelet values hk , gk and the values of the wavelet transform coefficients sj,k , d j,k make it possible to obtain an extrapolation based on the Haar-wavelet.

4 Solution of the Problem of Forecasting Self-similar Traffic Using the Method of Wavelet-Extrapolation Consider a “wavelet-extrapolation” using a specific example. Simulate self-similar traffic queuing system (QS) WB /M/1/K with the following parameters: – – – – –

λ is an intensity admission packages for service in the QS, λ = 150 pack/s, μ is a duration of the service packs, μ = 100 pack/s, K is a packet queue length, K = 1000 packages, Hurst parameter H = 0,65, Weibull distribution parameters α ≈ 0,6 and β ≈ 2,58.

The simulation results for the self-similar traffic in the Simulink package in the area of Matlab for the given source data that are shown in Fig. 4, where N is the number of packets, t is the packet arrival time [9–13]. The received traffic in the [2100; 2500] ms segment, according to [9–13], is selfsimilar, has large-scale invariance, the presence of “bursts” of packet intensity and a long-term relation between the moments of their arrival. Consider self-similar traffic on the segment [2075; 2115] ms. Let self-similar traffic be described by some function f (x), defined on the considered segment. Consider the extrapolation of the self-similar traffic f (x), using the Haarwavelet function (Fig. 4) defined by its base function (mother wavelet) ψ(x) and scaling function ϕ(x) by formulas (16–17). A necessary condition for extrapolation using wavelet functions is its definition by the number of samples equal to N = 2jn , where jn ≥ 1 determines the maximum possible number of scaling levels [20]. In our case, jn = jmax = 4, N = 24 = 16. In the Table 1 shows the calculated values of the coefficients sj,k and d j,k, obtained according to (20), while the coefficients s0,k are the values of the function f k = f (x k ) at the interpolation nodes.

Different Extrapolation Methods in Problems of Forecasting

225

N 1500 1200 900 600 300

0 3 2.1×10

3

3

2.2×10

3

2.3×10

2.4×10

2.5×103 t, ms

Fig. 4. The simulation results of self-similar traffic for QS type WB /M/1/K. Table 1. Calculated values of coefficients sj,k and d j,k . k

j=1 sj,k dj,k

0 31,5 -29,7 j=2

1 124,5 -12,5

2 42,3 10,8

j=3 j=4

3 3,78 3,0 k sj,k dj,k

4 24,6 18,7 0 110,3 -65,7

5 57,5 -29,8 1 32,6 27,3 k sj,k dj,k

6 66,6 36,9 2 58,1 -23,3 0 0,1 54,9 k sj,k dj,k

7 25,6 1,4 3 65,2 29,0 1 87,2 -5,0 0 -0,1 5,0

Perform an extrapolation of self-similar traffic using formula (14) [20], using the Haar-wavelet, for simulated self-similar traffic in the [2100; 2115] ms segment, the mother wavelet ψ(x) and the scaling function ϕ(x) are defined by the formulas (16–17). The extrapolation results using the Haar wavelet were obtained in Mathcad 15 and are presented in Fig. 5. Using the original data, according to issue 3, we obtain an extrapolation of self-similar traffic on the [2100; 2115] ms segment using a cubic spline (Fig. 5). In general, the proposed method of extrapolation based on wavelet functions, according to the authors, has a number of advantages in comparison with known methods. It is quite easy to implement, it has a low resource consumption and error, and it can also be used to control traffic in real time. As can be seen from Fig. 6, the use of Haar-wavelet functions to extrapolate selfsimilar traffic, characterized by bursts of intensity and sharp increases in function, leads to a global smoothing of the signal. Local “bursts” at the time of sudden changes in traffic intensity have an extrapolation error of about 5%. Similar results are provided

226

I. Strelkovskaya et al.

N 400

2

1

200

) )

0

t, ms

- 200

- 400 2075

2080

2085

2090

2095

2100

2105

2110

2115

Fig. 5. The results of the extrapolation of self-similar traffic on the interval [2100; 2115] ms, 1) simulated self-similar traffic, 2) extrapolation of self-similar traffic using Haar-wavelet.

by the use of a cubic spline, which has an extrapolation error of about 10%. However, applying the wavelet transform does not interfere with the recovery of these peaks due to the scalability property. N 400 2 1

200

3

0

t, ms

- 200

- 400 2075

2080

2085

2090

2095

2100

2105

2110

2115

Fig. 6. The results of the extrapolation of self-similar traffic on the interval [2100; 2115] ms 1) simulated self-similar traffic, 2) extrapolation of self-similar traffic using cubic spline function, 3) extrapolation of self-similar traffic using Haar-wavelet

Different Extrapolation Methods in Problems of Forecasting

227

Based on the results of traffic prediction, taking into account the maximum values of the workloads of IoT network objects, practical recommendations can be given on the redistribution of traffic in the network and the use of various IoT applications to prevent network overload. This will balance the load on IoT network objects and increase the efficiency of using M2M and D2D network devices.

5 Conclusions 1. The solution of the problem of forecasting self-similar IoT traffic, obtained using the Simulink software package in Matlab, is considered. 2. The use of the wavelet extrapolation method for forecasting the characteristics of self-similar traffic is proposed. 3. A comparative analysis of the spline and wavelet extrapolations on a specific example. The results showed that the use of spline extrapolation is appropriate for selfsimilar traffic, which is characterized by smooth changes in intensity. On the contrary, for traffic that is characterized by the presence of frequent and rapid changes in intensity, it is advisable to use wavelet extrapolation. 4. The proposed wavelet extrapolation method has several advantages compared to other known methods, first of all, it is sufficient simplicity in implementation, low resource consumption and forecast accuracy, which can be improved through the use of various wavelet functions. 5. A comparison of the results of forecasting the characteristics of IoT traffic based on the extrapolation method using the Haar-wavelet and the cubic spline functions confirmed the feasibility of using wavelet functions. 6. The results of the prediction of self-similar traffic of IoT network objects and M2M devices make it possible to improve the forecast accuracy, ensure its scalability and use for various IoT applications, thereby avoiding network overloads and exceeding the standard values of QoE characteristics.

References 1. Evans, D.: The Internet of Things. How the next evolution of the Internet is changing everything, Cisco White Paper (2011). https://www.cisco.com/c/dam/en_us/about/ac79/docs/ innov/IoT_IBSG_0411FINAL.pdf 2. GPP Study on Scenarios and Requirements for Next Generation Access Technologies: ETSI TR 38.913, V14.3.0 (2017). https://www.etsi.org/deliver/etsi_tr/138900_138999/138913/14. 02.0060/tr_138913v140200p.pdf 3. Krylov, V.V., Samohvalova, S.S.: Theory of telegraphic and its applications. BXV-Petersburg, SPb (2005) 4. Sheluhin, O.I., Osin, A.V., Smolskii, S.M.: Self-Similarity and Fractals. Telecommunication Applications. Phismatlit, Moscow (2008) 5. Iqbal, M.F., Zahid, M., Habib, D., John, L.K.: Efficient prediction of network traffic for realtime applications. J. Comput. Netw. Commun. 2019, 1–11 (2019). https://doi.org/10.1155/ 2019/4067135

228

I. Strelkovskaya et al.

6. Khan, N.S., Ghani, S., Haider, S., Khan, N.S.: Real-time analysis of a sensor’s data for automated decision making in an IoT-based smart home. Sensors (Basel) 18(6), 1711 (2018). https://doi.org/10.3390/s18061711 7. Tang, F., Fadullah, Z.M., Mao, B., Kato, N.: An intelligent traffic load prediction-based adaptive channel assignment algorithm in SDN-IoT: a deep learning approach. IEEE Internet Things J. 5(6), 5141–5154 (2018). https://doi.org/10.1109/JIOT.2018.2838574 8. Yu, Y., Song, M.S., Fu, Y., Song, J.: Traffic prediction in 3G mobile networks based on multifractal exploration. Tsinghua Sci. Technol. 18(4), 398–405 (2013). https://doi.org/10. 1109/TST.2013.6574678 9. Strelkovskaya, I.V., Solovskaya, I.N.: Approximation of self-similar traffic by splinefunctions. In: Modern Problems of Radio Engineering, Telecommunications and Computer Science: Proceedings of the XIIIth International Conference (TSET 2016), pp. 132–135, Slavske (2016). https://doi.org/10.1109/tcset.2016.7451991 10. Strelkovskaya, I., Solovskaya, I., Severin, N., Paskalenko, S.: Spline approximation based restoration for self-similar traffic. East.-Eur. J. Enterp. Technol. 3/4(87), 45–50 (2017). https:// doi.org/10.15587/1729-4061.2017.102999 11. Strelkovskaya, I., Solovskaya, I., Makoganiuk, A.: Spline-extrapolation method in traffic forecasting in 5G networks. J. Telecommun. Inf. Technol. 3, 8–16 (2019). https://doi.org/10. 26636/jtit.2019.134719 12. Strelkovskaya, I., Solovskaya, I.: Using spline-extrapolation in the research of self-similar traffic characteristics. J. Electr. Eng. 70(4), 310–3016 (2019). https://doi.org/10.2478/jee2019-0061. ISSN 1335-3632, ISSN 1339-309X 13. Strelkovskaya, I., Solovskaya, I., Makoganiuk, A.: Forecasting self-similar traffic using cubic B-splines. In: 3rd IEEE International Conference Advanced Information and Communication Technologies 2019, pp. 153–156, Lviv (2019). https://doi.org/10.1109/AIACT.2019.8847761 14. Abry, P., Baraniuk, R., Flandrin, P., Riedi, R., Veitch, D.: Wavelet and multiscale analysis of network traffic. IEEE Signal Process. Mag. 19(3), 28–46 (2002) 15. Zang, Y., Ni, F., Feng, Z., Cui, S., Ding, Z.: Wavelet transform processing for cellular traffic prediction in machine learning networks. In: IEEE China Summit and International Conference on Signal and Information Processing, China (2015). https://doi.org/10.1109/ChinaSIP. 2015.7230444 16. Han, Y., Dezhi, L., Guo, Q., Kong D.: Self-similar traffic prediction scheme based on wavelet transform for satellite Internet services. Lecture Notes of the Institute for Computer Sciences, vol. 183, pp. 189–197 (2018). https://doi.org/10.1007/978-3-319-52730-7_19 17. Asars, A., Grab, E., Petersons, A.: Analysis of wavelet estimation of self-similar traffic parameters in the Simulink model. Autom. Control Comput. Sci. 47(3), 132–138 (2013). https:// doi.org/10.1109/ChinaSIP.2015.7230444 18. Zavyalov, Yu.S., Kvasov, B.I., Miroshnichenko, V.L: Methods of spline functions. Science, Moscow (1980) 19. Popovsky, V.V., Strelkovskaya, I.V.: Accuracy of filtration procedures, extrapolation and interpolation of random processes. Probl. Telecommun. 1(3), 3–10 (2011) 20. Dremin, I.M., Ivanov, O.V., Nechitaylo, V.A.: Wavelets and their use. Successes Phys. Sci. 171(5), 465–501 (2011) 21. Strelkovskaya, I., Lysiuk, O., Paskalenko, S.: Application of different kinds of approximation in signals restoration. Probl. Infocommun. Sci. Technol. 2015, 177–180 (2015). https://doi. org/10.1109/INFOCOMMST.2015.7357306 22. Strelkovskaya, I.V., Lysiuk, O., Zolotukhin, R.: Comparative analysis of signals restoration by different kinds of approximation. In: 3rd International Conference on Applied Innovations in IT, ICAIIT-2015, vol. 3, no. 1, pp. 29–35 (2015) 23. Dobeshi, I.: Ten lectures on wavelets. SIC Regular and Chaotic Dynamics (2001)

Methods for Calculating the Performance Indicators of IP Multimedia Subsystem (IMS) Oleksandr Romanov(B) , Mykolaiy Nesterenko(B) , Leonid Veres(B) Roman Kamarali(B) , and Ivan Saychenko(B)

,

National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Peremoga Avenue 37, Kyiv 03056, Ukraine [email protected], [email protected], [email protected], [email protected], [email protected]

Abstract. Today, Ukraine is actively working on solving the problems of convergence of networks of various technologies, ensuring their compatibility in management, signaling and data traffic, providing users with modern services with specified QoS indicators. In the process of implementing 4G and 4.5G technologies in the networks of mobile operators, it turned out that the voice traffic service is possible only when using a core network with switching channels of 2G and 3G technologies. And in order for voice traffic to be serviced in batch mode, it is necessary to bring the network of the mobile operator in accordance with the requirements of the IMS architecture. The implementation of IMS provides a solution to two main tasks. The first is the solution to the problem of interaction between networks of various technologies at the level of signal flows. The second task is to ensure the interaction of networks of various technologies at the level of data flows. These tasks are solved at the control level, the main functional elements of which are the functional blocks CSCF, P-CSCF, I-CSCF, S-CSCF. When designing the IMS management level, it is necessary to determine the amount of traffic that will be served by these CSCF functional elements. Then, it is necessary to calculate the required performance of the elements for serving signal traffic and data streams. At the same time, it is very important to provide the required reliability indicators for the IMS management level. The article analyzes the functioning of the IMS core in solving problems of combining signaling traffic and data networks of various technologies. The list of the main elements involved in the process of solving management problems is determined. Models are proposed that allow calculating the amount of incoming traffic from heterogeneous networks and the amount of traffic after it is converted to the standard form of IMS architecture. To ensure the indicated reliability indicators of IMS, the use of methods of reservation of functional elements of the control level is provided. Models have been developed and analytical expressions have been obtained for calculating the values of reliability indicators using various reservation methods. Keywords: IP Multimedia Subsystem · IMS · CSCF · P-CSCF · I-CSCF · S-CSCF · CSCF reliability · CSCF reservation methods · CSCF separate reservation · CSCF general reservation © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 229–256, 2021. https://doi.org/10.1007/978-3-030-58359-0_13

230

O. Romanov et al.

1 Introduction Today in Ukraine, as in many countries of the world, active work is being done to solve the problems of convergence of networks of various technologies, to ensure their compatibility at the control levels, signaling and traffic flows, to provide users with modern services with specified QoS indicators. Initially, the development of telecommunication networks was carried out within the framework of the concept of NGN (Next Generation Network), which was provided for the solution of the following main tasks: – Creation of a common telecommunication space, which would ensure the interaction of networks regardless of the technologies used in them; – the allocation of levels within which technology development could be carried out by specialists in various fields of knowledge independently of each other; – harmonious development of telecommunications, in which networks using efficient technologies would occupy a larger segment of the market, and networks using outdated technologies would gradually die out; – rapid implementation of new services with minimal cost; – independence of the service functions of the network from the transport technologies used in it; Research in the field of NGN has shown that in promising networks a clear functional differentiation of levels should be implemented [1, 2]. At the same time, the four-level architecture of NGN was most widely used: 1. 2. 3. 4.

The level of service. The level of management. The level of the backbone transport network. The level of access networks.

Simplified model of NGN network architecture can be represented in the following form (Fig. 1). Application Servers Level of service and apps

Level of transport network Level of aggregation, control and signaling Level of access networks

Fig. 1. Simplified model of NGN network architecture

We briefly consider the tasks that were solved at each level.

Methods for Calculating the Performance Indicators of IMS

231

The level of services provides: 1. Accessibility of services to all users, regardless of their location, access method and type of transport network that serves them. 2. Interaction with the core transport network through open interfaces. 3. The ability to develop software services and applications, focusing only on internal standard interactions with open interfaces. 4. Introduction of new services without interfering with the functioning of other levels. As a mechanism for quickly and flexibly deploying services depending on the individual needs of users, the concept of OSA (Open Services Access) was used [3]. One of the options for implementing the OSA concept in relation to telecommunication services was the creation of an open Application Programming Interface (API) by the Parlay Group consortium to manage network resources and access network information. Figure 2 shows the Parlay architecture, which is one of the practical implementations of the OSA concept.

Fig. 2. Architecture OSA/Parlay

As shown in Fig. 2, different communication networks have different network elements that provide the required functions: – The second generation mobile network includes SGSN (Serving GPRS Support Node) and MSC (Mobile Switching Center); – The public switched telephone network includes an SSP (Service Switching Point) service switch; – The IMS network includes S-CSCF (Serving Call Session Control Function), etc.

232

O. Romanov et al.

Each of these elements is connected to the gateway via its own protocol, and the task of the OSA/Parlay gateway is to reduce all protocols to common APIs. Then you can write applications without taking into account the features of underlying networks, and you should only strictly adhere to the APIs. The management level is designed to solve the following problems: 1. Ensuring the interaction of networks with switching channels and networks with packet switching. 2. Combining traffic flows of networks with various technologies into a common homogeneous flow. 3. Ensuring the interaction of various alarm systems, in particular, alarm systems based on the SIP protocol and SS 7. 4. Managing the flow of voice messages, analogue telephony and IP-telephony. 5. Routing message flows. 6. Transformation and ensuring the interaction of various alarm systems; 7. Formation of information for billing systems, etc. The main control element in the NGN network is the software switch - Softswitch. In addition, Media Gateways (MGW) can be used that convert information flows into access network format and direct them to the desired channels. These include, in particular, DSLAM and IP/QAM gateways installed at the borders of access networks. One SoftSwitch can control many media gateways, the distance between which can reach thousands of kilometers. It should be noted that SoftSwitch from different manufacturers turned out to be incompatible with each other in terms of functions and interfaces. This deficiency led to the need for further development of the concept of NGN. The tasks of the transport layer are switching and transparent transmission of user information. The NGN transport layer is considered as a layer whose components are the access network and the core network. The core network is a universal network that implements the functions of transport and switching. In accordance with these functions, the core network can be represented in the form of three levels (Fig. 3):

Packet (frame) switching technology (IP, ATM, Frame Relay, MPLS√ )

Network level

Technology of tract formation (SONET/SDH, Ethernet √ )

Channel level

The signal transmission environment (WDM optical fiber, DWDM, radio relay line...)

Physical level

Fig. 3. Base transport network model

Methods for Calculating the Performance Indicators of IMS

233

– level of packet switching technology; – the level of technology for the formation of channels and paths for the transmission of information; – level of signal transmission medium. The lower level of the model is the signal transmission medium. This level can be implemented on trunk cables, cables with optical fibers or on microwave lines. Technologies xWDM, Carrier Ethernet, SDH, as well as their combination can be used as technologies for the formation of channels and paths of information transfer [4–7]. When choosing the technological basis for switching, the most promising is the IP/MPLS technology. This is due to the fact that: – Using IP/MPLS technology in an Ethernet environment can increase scalability and quality of service to the level required for transport networks. – The MPLS RSVP-TE specifications allow you to provide specified indicators of throughput and quality of service. Thus, the concept of NGN allowed to solve the following problems: 1. When introducing new services for the operator, there is no need to adapt them to the capabilities of the transport structure. 2. At the management level, you can integrate the services of various suppliers. 3. The use of unified interfaces allows you to connect the transport core with both heterogeneous access networks and server equipment at the level of services and applications. 4. Integration of fixed and mobile communication networks, as a result of which the task of providing subscriber mobility is solved. 5. Identification of subscribers in all access segments and the creation of a single subscriber profile that ensures the functioning of the accounting and billing system. 6. Providing end-to-end quality of service requirements for all types of services, including streaming and multimedia. 7. Maintaining the ability to provide services of traditional public networks, transmitting technical alerts and operational alerts. Despite the fact that NGN networks solved many problems in the development of telecommunication networks, they had one big drawback. This drawback is associated with SoftSwitch equipment, which was the main element of the control level. It turned out that the SoftSwitch equipment of various manufacturers are incompatible with each other either in function or in interface. This shortcoming led to the need to develop a new concept, which was called the concept of IMS (IP Multimedia Subsystem). Changes in architecture during the transition from NGN networks to IMS networks are presented in Fig. 4 [8–11]. An analysis of Fig. 4 shows that the NGN and IMS architectures are similar to each other. The boundaries of logical levels are the same. At the same time, the level of services and applications remained unchanged. The access

234

O. Romanov et al.

network layer and the transport layer of the NGN network in the IMS network have been combined into one layer. It is called the transport layer of the IMS network. NGN network

IMS network

Level of services and applications IVR-сервер

IP server

Application Servers (AS)

Management level

S-CSCF

SoftSwitch I-CSCF

P-CSCF

Transport level

Transport level TDM

IP

IP Network

Access level

Fig. 4. Changes in architecture during the transition from NGN

The main changes have occurred in the principles of building the level of management. In the NGN network, SoftSwitch, called the software-configurable switch, was used at the control level. It was a monolithic physical device. The principle of its construction was in many ways similar to the automatic telephone exchange (PBX) device. SoftSwitch had many different kits, each of which provided interfacing with a network of a certain technology. The low level of standardization of interfaces, functions and the list of tasks to be solved has led to the fact that the SoftSwitch equipment of various manufacturers turned out to be incompatible with each other. The transition to IMS is characterized by a completely new approach to the principle of implementation of the management level. Instead of a monolithic physical device, SoftSwitch began to use a set of functional blocks at the control level. These are not specific physical elements, but a multiservice platform with a set of proxy servers. Each server provides: – management of the provision of services of a certain type; – its own interface for interacting with the transport network; – all servers are connected to a single database, which contains all the information about traffic and received services; – the database has an external interface that allows third-party suppliers to obtain the information necessary for mutual settlements. The functions of the control level can be software implemented on any server hardware. Thus, the management level has become independent of the equipment manufacturer.

Methods for Calculating the Performance Indicators of IMS

235

2 Analysis of the Principles of Construction and Functions of the Basic Elements of IMS IMS has three levels: the transport level, the management level, and the service level (Fig. 5). Consider the purpose and functions of the basic elements of the IMS architecture levels. The IMS architecture is defined in the 3GPP (3rd Generation Partnership Project) of the European Institute of Communication Standards [12–16]. Signal informaon

Data transfer informaon (audio, video)

Applicaon level AS

SCIM

HSS S-CSCF

BGCF

SGW

P-CSCF I-CSCF

MRF

MGCF

Level of control NASS

A-RAGS

PDF

GW PSTN

PSTN

MRCF

MRFP MGW DSLAM

BGF

WAG

PDG

RAN

TrGW

GGSN

IPv4

I-BGF Transport level

SGSN

IPv6

Fig. 5. Core functional blocks of the IMS architecture

At the transport level, the following main tasks are being solved: 1. 2. 3. 4.

Providing users access to the transport network. Creating a signal transmission medium. Formation of channels and paths for information transfer. Switching and distribution of information.

To solve these tasks, the transport elements include the following basic elements: MGW (Media Gate Way), I-BGF (Interconnect Border Gateway Function), GGSN (Gateway GPRS Support Node), RAN (Radio Access Network), PDG (Packet Data Gateway), WAG (Wireless Access Gateway), A-BGF/BAS, DSLAM (Digital Subscriber Line Access Multiplexer) Consider the purpose and function of the basic elements. 1. MGW (Media Gate Way) - a transport gateway that is designed to convert E1–E4 circuits with circuit-switched (PSTN) flows to RTP packets of IP networks and vice versa. 2. I-BGF (Interconnect Border Gateway Function) is an edge gateway that enables the interworking of IPv4 and IPv6 networks.

236

O. Romanov et al.

3. RAN (Radio Access Network) - radio access network equipment. Provides interaction of mobile subscribers with a stationary component of cellular networks; 4. GGSN (Gateway GPRS Support Node) - gateway to the interaction of wireless access networks with IMS architecture. It ensures the coordination of the connection parameters of mobile users with the IMS architecture, both for signaling and for transport flow. 5. WAG (Wireless Access Gateway) - wireless access gateway. Provides access to user equipment WIFI to a fixed network. 6. PDG (Packet Data Gateway) - packet gateway. Provides interoperability of WIFI wireless access equipment with a packet network. 7. DSLAM (Digital Subscriber Line Access Multiplexer) - Broadband access equipment. It is intended for connection of subscribers with the equipment of Internet service providers and IPTV. 8. BGF/BAS - a gateway that provides the coordination of broadband access networks with IMS architecture equipment for signaling and the type of transport stream. 9. TrGW - provides the negotiation of a circuit switched domain (PSTN) and an IMS network via a call system At the management level, tasks are being accomplished to harmonize signaling networks, distribute information and organize the interaction of telecommunication networks with various technologies. It provides the solution of the following main tasks: 1. 2. 3. 4.

Use of SIP as the basic signaling protocol. Bringing networks of different technologies to a single transport protocol IP. Using a single scenario of user service, both in fixed and mobile networks. Creating a user profile in the home HSS database and importing this data into guest databases when moving subscribers’ other domains. 5. Creation of unified principles of authorization, identification and tariffing for networks of various technologies. 6. Providing access to any service, regardless of its location. To solve these tasks at the management level, there are the following basic elements: CSCF (Call Session Control Function), P-CSCF (Proxy Call Session Control Function), I-CSCF (Interrogating CSCF), S-CSCF (Serving-CSCF), BGCF (Break Gate Control Function), MGCF (Media Gateways Control Function), SGW (Signaling Gateway), RACS (The Resource and Access Control), PDF (Policy Decision Function), NASS (Network Attachment Subsystem). Consider the purpose and function of the basic elements. The CSCF function block performs all management tasks in the IMS architecture. These tasks are divided between three functional blocks: 1. The first functional block is the P-CSCF. It is designed to interact with all kinds of end-user terminals, create and record a user profile in the home HSS database. 2. The second functional unit is Interrogating CSCF. It is designed to interact with other IMS domains. For example, one of its main tasks is to determine the location of a subscriber if its HSS is in another IMS domain.

Methods for Calculating the Performance Indicators of IMS

237

At the management level, tasks are being accomplished to harmonize signaling networks, distribute information and organize the interaction of telecommunication networks with various technologies. It provides the solution of the following main tasks: 1. 2. 3. 4. 5.

Use of SIP as the basic signaling protocol. Bringing networks of different technologies to a single transport protocol. Using a single scenario of the user service, both in fixed and mobile networks. Creating a user profile in the home. Creation of unified principles of authorization, identification and tariffing for networks of various technologies. 6. Providing access to any service, regardless of its location. To solve these tasks at the management level, there are the following basic elements: CSCF (Call Session Control Function), P-CSCF (Proxy Call Session Control Function), I-CSCF (Interrogating CSCF), S-CSCF (Serving-CSCF), BGCF (Break Gate Control Function), MGCF (Media Gateways Control Function), SGW (Signaling Gateway), RACS (The Resource and Access Control), PDF (Policy Decision Function), NASS (Network Attachment Subsystem). Consider the purpose and function of the basic elements. The CSCF function block performs all management tasks in the IMS architecture. These tasks are divided between three functional blocks: 1. The P-CSCF functional element provides interaction with all types of end-user terminals, and also performs filtering and protection of the operator’s network from harmful information. 2. The I-CSCF functional entity provides interworking with other IMS domains. For example, one of its tasks is to locate a subscriber if its HSS is in a different IMS domain. 3. The S-CSCF functional entity processes all signaling SIP messages and also provides subscribers’ access to application servers in accordance with the service profile contained in the home server of the HSS subscriber.

3 The Real-Time Traffic Service Protocols in IMS The main type of real-time traffic with the highest quality of service requirements is voice traffic in the IP telephony system. Therefore, we consider the main components of IMS that provide voice messaging in accordance with existing standards [17–20]. At the same time, we will pay attention to the features of information exchange between end users (Fig. 6). As shown in Fig. 6, paths signaling traffic and media streams are separated not only logically but also physically. That is, media streams are transmitted at the transport level (core network). A signaling and control are carried out at higher levels, and their management elements can be territorially separated.

238

O. Romanov et al.

Application layer

(MCFC)

(Application servers)

HSS

IP Media Server / Media Resource Function (MRF )

CSCF

(Call Session Control Function)

Session Control Layer

SIP-server SIP-server IMS: IP Mutlimedia Subsystem

Transport Layer

Core network (IP/MPLS)

UTRAN/ EDGE/ GRRS/ Cable/ xDSL, WiFi, etc.

SIP client

Access Layer

(MRFP)

Signaling Flow Media Flow

SIP client

UTRAN : UMTS Terrestrial Radio Access Network GPRS : General Packet Radio Service EDGE : Enhance Data Rates for Global Evolution MRFC : Multimedia Resource Function Controller MRFP : Multimedia Resource Function Processor HSS : Home Subscriber Server

Fig. 6. Network architecture and service architecture IMS

It is also necessary to take into account the protocols of VoIP networks, which were deployed earlier and successfully used by different manufacturers of network equipment: networks H.323; networks MGCP; networks H.248/MEGACO; SIP networks. Consider each technology in more detail. The structure of the network H.323 (Fig. 7) consists of four main functional components.

Terminal Н.323

Terminal Н.323

Multipoint Conference Unit (MCU)

Gatekeeper

Router Gateway WAN (Internet)

ISDN

GSM-network

PSTN

Private Branch Exchange (PBX)

Fig. 7. (2). Architecture H.323 network

Methods for Calculating the Performance Indicators of IMS

239

Assignment of network elements: 1. H.323 Terminal—a subscriber device that can communicate (voice, video, etc.) with other terminals, gateways or multipoint conference unit. 2. H.323 Gateway—is the central element of today’s IP telephony. This device provides the interaction of elements of the public telephone network with IP network. 3. H.323 Gatekeeper—is the control element, the „intelligence” of the H.323 network. It provides its scaling, centralization of management functions and settings, as well as the transfer of telephone prefixes and identifiers (H.323 ID) to IP addresses of gateways or H.323 terminals. Gatekeeper is responsible for managing access (access control) when registering gateways and terminals, authorization of calls, bandwidth management and call routing. 4. Multipoint Conference Unit (MCU)—manages the conduct of multi-user conferences, coordinates the connection parameters of all participants in a centralized, decentralized or combined conference. In the most general form, the H.323 connection scenario looks like a sequence of subsequent steps (Fig. 8). First, to establish a connection, the terminal finds the Gatekeeper where it registers on it in accordance with the RAS protocol. Then the procedure for installing the signal channel through the protocols RAS and H.225 is executed. The next step is to reconcile the hardware parameters, exchange information about its functionality and open the logical channels by the protocol H.245. Only after this is the transfer of media traffic by protocols RTP/RTCP. After the closing session, the connection closes in the reverse order. Request for access Confirmation of access

H.225 (UDP) RAS

Gatekeeper Gateway or Terminal Н.323

Connection request

H.225 (TCP) Q.931

The connection is established Gateway Selection of options media-channel Open media-channel

H.245 (UDP / TCP)

Confirmation of opening media-channel

RTP-flow RTP-flow RTP-flow

Media-traffic (UDP)

Fig. 8. Connection scenario for H.323 protocol

Protocol MGCP (Media Gateway Control Protocol) has been proposed by the MEGACO Working Group (IETF). The basic idea behind MGCP is that the control

240

O. Romanov et al.

of the signaling is concentrated in the central control unit and completely separate from the management of media streams. These flows are processed by the gateway, subscriber terminals that perform commands coming from the control device [21, 22]. In the architecture of the MGCP network, you can distinguish two main functional components: 1. Media Gateway (MG) or IP-phone—converts the voice information from a fixedspeed telephone network to a format suitable for transmission over IP networks: encoding and decoding of media traffic. 2. Call control device: can be called the Signaling Controller/Call Agent (CA); Media Gateway Controller (MGC); or software-hardware controller (Softswitch, SS). Sometimes the signaling controller is represented as two components: – Call Agent (controller) who performs gateways management functions; – Signaling Gateway, which transfer signaling traffic from the PSTN side to the gateway control unit and the signaling traffic in reverse direction. In Fig. 9 shows the architecture and main components of the MGCP network. Call Agent

Media Gateway

Signaling Gateway

Telephone line

Private Branch eXchange (PBX)

SS7 MGCP IP network

Е1 RTP

Media Gateway

Media Gateway

Private Automatic Branch eXchange (PABX)

Fig. 9. Architecture MGCP network

The signaling controller (CA) perceives the network as a set of two logical elements - endpoints and connections between them. Devices can be physical (for example: IP telephones, or gateway lines) or virtual (for example, soft port numbers of voice message servers). Element management and devices communication are performed by sending commands as text (ASCII) messages over UDP, also can use the SDP protocol. The simplest installation scenario is establishing a connection in the MGCP network. The user of the phone that is connected to the MGCP gateway takes off the handset. The

Methods for Calculating the Performance Indicators of IMS

241

gateway reports to the controller (Call Agent) about this event. The Call Agent gives the Gateway command to include a dial-ton signal in the telephone line. Now the user is hearing a continuous beep in the handset. Then the phone number is dialed, it is also a sequence of events for the controller. By analyzing these events, the Call Agent establishes a connection to another subscriber in the IP network or telephone network. The MEGACO/H.248 protocol is an evolutionary development of the MGCP protocol, which has the basic conceptual properties of its predecessor [23–25]. Namely, the basis of the protocol MEGACO/H.248 includes the principle of decomposition of the functions gateway. That is, the gateway is divided into separate functional blocks (Media Gateway; Signaling Gateway; Media Gateway Controller). The whole “intelligence” of the architecture is in Media Gateway Controller. The following transport protocols can be used for the transport of MEGACO/H.248 signaling messages: UDP, TCP, SCTP (Stream Control Transport Protocol), and ATM technology. UDP protocol support is a mandatory requirement for the MGC. The TCP protocol must be supported by both the controller and the gateway. Messages from the MEGACO/H.248 protocol can be encoded in two ways. The IETF has proposed a text encoding method for signaling and SDP protocol for describing communication sessions. ITU-T has provided a way for signaling information to be provided by means of abstract synthesis - ASN.1, and for the description of communication sessions, recommends a special tool “Tag - length - value” (TLV). The MGC controller must support two methods of encoding, and the Media Gateway is one of them. MEGACO is an “internal” protocol that operates between the distributed gateway functional blocks, between the MGC and the MG. The principle of this protocol is master (MGC)/slave (MG). The SIP protocol (Session Initiation Protocol)—sets up and completes multimedia sessions as well as sessions for sharing video content and text, collaborative work on applications (RFC 2543) [26, 27]. SIP was developed by analogy with HTTP and SMTP (client-server protocol). All SIP headers are transmitted in ASCII text format. SIP allows you to use logical addressing (URL) based on TCP or UDP (for example: sip: [email protected]). To implement additional capabilities, SIP uses: – SDP (Session Description Protocol, RFC 2327), protocol for the coordination of communication parameters (types of codecs, UDP port numbers, etc.). – SDP provides the change of the „on the fly” session parameters during the session. The transfer of SDP messages is based on the Session Announcement Protocol (SAP, RFC 2974). The architecture of SIP (Fig. 10) consists of the following components: 1. SIP client (User Agent)—can be presented as a hardware-software device (IPphone, gateway or user terminal) or software application. The main functions of this component are initiating and completing calls.

242

O. Romanov et al. Application servers

e-mail

3pcc LDAP

SQL

CPL

XML

SIP server: Proxy server; Registrar server; Redirect server. SIP server

SIP

SIP server SIP server SIP server

SIP server SIP server

Public Switched Telephone Network (PSTN)

SIP clients (User Agents)

Signaling flow (СAS / PRI)

RTP sip phone

sip phone Private Automatic Branch eXchange (PABX)

Fig. 10. Architecture SIP.

2. SIP Proxy server—manages call routing and application work. The proxy server can not initiate or terminate calls. 3. SIP Redirect server—redirects calls according to specified conditions. 4. SIP registration server (registrar/location)—registers users and maintains the database of user names for their addresses, telephone numbers, etc. Another important component of real SIP networks is the Back-to-Back User Agent (B2BUA). This is a peculiar server, which is a combination of two SIP clients and therefore capable of initiating and completing calls. In the most common form, the scenario for installing a SIP connection with a proxy server is shown in Fig. 11.

SIP server Public Switched Telephone Network (PSTN)

Public Switched Telephone Network (PSTN)

SIP network Gateway A

Gateway B INVITE 100 Trying

Signaling traffic SIP and SDP (UDP / TCP)

INVITE 100 Trying 180 Ringing

180 Ringing

200 OK

200 OK ACK

Selection of options media channel Connection establishment

ACK

RTCP Media traffic (UDP)

RTP flow

Media channels

Fig. 11. Connection scenario for SIP protocol

Methods for Calculating the Performance Indicators of IMS

243

The subscriber sends a request for a connection to the proxy server using the INVITE (service request) message. The proxy server returns a Trying message and sends the INVITE message to the calling party. In case of availability of the subscriber’s terminal to which the call is made, the gateway corresponds to the Ringing message, which the proxy sends to the calling party. After a subscriber takes a phone call, gateway sends a message OK to the subscriber who made the call. After this, the subscriber who made the call, the equipment returns an ACK confirmation message. From now on, the RTP/RTCP protocols begin to exchange media traffic. The party, wishing to terminate the connection, sends a BYE message, and after receiving of the confirmation message OK, connection is broken. To find out the features of the interconnection of subscribers, consider the basic scenarios for establishing a connection using segments of networks with different technologies that are integrated into the IMS domain. Figure 11 shows the main phases signaling message exchange between mobile communication clients. Thus, when establishing a connection between subscribers, the SIP signaling protocol is used. Contact points IMS for SIP Users (SIP-UA) are Proxies-CSCF. Serving-CSCF (S-CSCF) is SIP server that connection and controls communication sessions. If the subscribers belong to other IMS domains (or there are several Serving-CSCFs in the same domain), another SIP server is used to locate the required S-CSCF—Interrogating CSCF (I-CSCF). Real-time Transport Protocol (RTP) and RealTime Transport Control Protocol (RTCP) are used for transmission of voice and video streams in IP networks. In turn, GGSN (GPRS Gateway Service Node) is a gateway (router) between a network for GPRS data transmission and an external Packet Data Networks (PDNs): Internet, corporate networks Intranet, other GPRS systems. The main task of GGSN is the routing of data between GPRS Core network and external IP-networks. In the case of convergence PSTN networks and IMS, the principle of signaling message exchange is shown in Fig. 12. This figure shows that in addition to SIP, the following signaling protocols are used in the connection process: SIGTRAN, SS7, MEGACO.

S-CSCF

P-CSCF

IMS

SIP signaling

P-CSCF

SIP signaling

UMTC network

User 1 GPRS Gateway Service Node (GGSN)

Voice, Video

GPRS Gateway Service Node (GGSN)

User 2

Fig. 12. SIP signaling paths and media streams in the same IMS domain for mobile networks

244

O. Romanov et al.

This indicates that it is necessary to take into account the transformation of signaling messages and increasing signaling traffic in network. That is, the most complex transformations of media streams and additional processing of signaling occurs at the boundary of the converged network and the IMS domain (Fig. 13).

Application server SIP

Application SIP server

SIP Media Resource Function (MRF )

Application server

SIP DIAMETER

Signaling Gateway Public Switched Telephone Network (PSTN)

SIP

SIP

SIGTRAN / Н 248

SS7

S-CSCF

SIP

SIP

Н 248

Softswitch

SIP

P-CSCF SIP

DIAMETER

HSS

I-CSCF Trunking Gateway

IMS

Fig. 13. The general structure of the core of the IMS network

4 Model and Method for Determining IMS Management Level Performance Requirements The analysis of the IMS architecture, the composition of its elements, protocols and the functioning process allows us to identify the main elements involved in the process of solving management problems. Based on the results obtained, IMS system models can be proposed to determine bandwidth and reliability requirements. First, consider a model that allows you to determine the bandwidth requirements of the IMS core (Fig. 14). The model contains only the basic elements that are directly involved in servicing traffic. In addition, to reduce the size and complexity of the task, the network is divided into three segments. Segment 1 displays the loading of network elements that communicate, based on the SIP protocol. Segment 2 displays the load generated by the circuit-switched public network. Segment 3 displays the load that is created when the HSS database interacts with the level of services and applications. Let us analyze the main tasks solved by the CSCF function block in the process of servicing incoming applications, and we will give suggestions on how to calculate the necessary transport resource in the core of the IMS architecture. To this end, we introduce the notation and define the parameters of each of the three blocks of functions S-CSCF, P-CSCF and I-CSCF, which form the basis of the CSCF function block. The S-CSCF function block is a SIP server that manages the communication session. To do this, he receives all the information about the connection and the necessary services from other network elements.

Methods for Calculating the Performance Indicators of IMS

245

Fig. 14. Elements of the IMS system that are directly involved in the traffic service.

We introduce the following notation: Denote the average number of SIP messages for one call between – – – –

MGCF i S-CSCF – Nsip1 , MRF i S-CSCF – Nsip2 , AS i S-CSCF – Nsip3 , I-CSCF i S-CSCF – Nsip4 ,

The average length of a SIP message in bytes – Lsip ; X% - the percentage of calls that require access to the MRF server when servicing; Y%. Percentage of calls that require access to AS application servers; Vss−s−cscf – The transport resource between the MGCF and the S–CSCF that is required for SIP message exchange during call handling; Vas−s−cscf – A transport resource between application servers (AS) and S–CSCF, which is required for SIP message exchange during call handling; Vmrf−s−cscf – The transport resource between MRF and S–CSCF, which is required for SIP message exchange during call handling; Vi−cscf−s−cscf – The transport resource between the I–CSCF and the S–CSCF that is required for SIP message exchange during call handling; Vs−cscf – The common transport resource of the S-CSCF that is required for the SIP message exchange during call handling. Then the total required transport resource will be equal to the total transport resource of the interaction of the S-CSCF function block with other IMS architecture elements:

246

O. Romanov et al.

To translate the expression from “bytes per hour” kbit per second”, you must enter the following coefficient:

The functional element of the I-CSCF, as well as the S-CSCF, is involved in servicing calls of disparate networks such as PSTN. Let’s estimate a transport resource is required for I-CSCF interaction with network elements. We introduce the notation: Vi−cscf - Traffic, which is sent to the I-CSCF; Vmgcf−i−cscf - traffic between the MGCF and the I-CSCF; Vp−cscf−i−cscf - transfer between P-CSCF and I-CSCF; Then the total traffic will be: Vi−cscf = Vmgcf−i−cscf + Vi−cscf−s−cscf + Vp−cscf−i−cscf Value Vp−cscf −i−cscf Was determined earlier, and Vmgcf−i−cscf Is calculated as follows: (Bit/s). We define the necessary transport resource for the P-CSCF element to provide signal exchange with the function block I-CSCF and PDF. To do this, we introduce the notation: – P-CSCF i I-CSCF – Nsip6 . – P-CSCF i PDF – Nsip7 . In addition, we denote: Vp−cscf - the common transport resource of the I-CSCF that is required for SIP message exchange during call handling, Vpdf−p−cscf - transport resource between the MGCF and the I-CSCF; Then the general transport resource: Vp−cscf = Vpdf−p−cscf + Vp−cscf−i−cscf Value Vp−cscf−i−cscf Calculated earlier, a Vpdf−p−cscf Is calculated by the formula: Thus, using this technique, you can determine the amount of load that functions between elements of the IMS architecture. Further, by setting the quality of service indicators, the performance requirements for the branches of the transport network are determined.

5 Models and Method for Calculating the Reliability of the IMS Management Level The structure of the IMS system, shown in Fig. 1, is rather complicated. It does not allow to apply immediately the mathematical apparatus of the theory of reliability of

Methods for Calculating the Performance Indicators of IMS

247

systems and calculate the values of reliability indexes of the IMS control level. In order to solve this problem, it is necessary to simplify the structure of Fig. 5, leaving only those elements that reflect the essence of the reliability of the system. We produce a series of consecutive steps of transition from a real system to its model: 1. Since the purpose of the work is to analyze methods for improving the reliability of the functioning of networks of mobile operators, we remove from the architecture of Fig. 1 the elements of stationary networks. 2. Simplify the application layer by presenting it as a single primary server for TAS services. 3. IMS control level is represented as a sequential chain of elements P-CSCF, I-CSCF, S-CSCF. Considering the above, the architecture of the system of Fig. 1 can be represented in a simplified form of Fig. 15.

Fig. 15. Simplified IMS architecture with highlighted main functional elements of management and application levels

However, the system in Fig. 8 has redundant elements that have little effect on the reliability indexes of the control level. Therefore, we will designate all elements of mobile communication networks of various technologies by the functional block “IMS transport network level” and then for analyzing the reliability of operation, we can use the system architecture, which is shown in Fig. 16.

248

O. Romanov et al.

Fig. 16. Simplified IMS architecture for solving reliability analysis problems

On the basis of this simplified IMS architecture, it is possible to propose a model that will have the form shown in Fig. 4. The following notation is introduced in Fig. 17:

Fig. 17. Model for calculating the IMS management level

– – – – – –

Wtransport level – the probability of failure-free of the transport network level; Waplicat. level – the probability of failure-free operation of the application level IMS; Wend user equip. – probability of end user equipment failure; ωP−CSCF – probability of failure-free operation of the functional element P-CSCF; ωI−CSCF – probability of failure-free operation of the functional element of I-CSCF. ωS−CSCF – probability of failure-free operation of the functional element of S-CSCF.

In the process of analyzing the reliability of the IMS management level, we will use the following limitations: – probability of failure-free operation Wtransport equals 1, i.e.

level

of the transport level network IMS

Wtransport level = 1 – probability of failure-free operation Waplicat. level of the application level network IMS equals 1, i.e. Waplicat. level = 1 – probability of failure-free operation Wend 1, i.e. Wend

user equip.

user equip.

of the end user equipment equals

=1

In addition, in the process of analyzing the reliability indexes we will use the following parameters:

Methods for Calculating the Performance Indicators of IMS

249

– PP−CSCF = 1 − ωP−CSCF – probability of failure-free operation of the functional element of P-CSCF; – PI−CSCF = 1 − ωI−CSCF – probability of failure-free operation of the functional element I-CSCF; – PS−CSCF = 1 − ωS−CSCF – probability of failure-free operation of the functional element S-CSCF. Calculate the reliability of the system in Fig. 17. This system consists of seriesconnected functional elements with reliability indexes Wtransport level , Waplicat. level , ωP−CSCF , ωI−CSCF , ωS−CSCF Wend user equip. . Probability of failure-free operation WP−CSCF_I−CSCF_S−CSCF such a system can be calculated as follows: WP−CSCF_I−CSCF_S−CSCF = ωP−CSCF ∗ ωI−CSCF ∗ ωS−CSCF And the probability of failure PP−CSCF_I−CSCF_S−CSCF equals PP−CSCF_I−CSCF_S−CSCF = 1 − ωP−CSCF ∗ ωI−CSCF ∗ ωS−CSCF So as indicators of reliability Wtransport level , Waplicat. level , Wend user equip. equal to one, then in further studies we will consider only indexes of the reliability of the functional elements P-CSCF, I-CSCF, S-CSCF.

6 Analysis of Methods to Improve the Reliability of the IMS Management Level There are many ways to improve system reliability by introducing redundancy. This article will cover two methods: 1. General redundancy, in which the entire system as a whole is backed up. 2. Separate reservation, in which each element has its own separate reserve. The general redundancy method assumes the reservation of the entire system as a whole. In general, for a system consisting of n elements with values of reliability indexes (ω11 , ω12 , . . . , ω1i . . . ω1n ), and assuming m is multiple redundancy, the model for calculating reliability can be represented as Fig. 18. Figure 18 presents the scheme of permanent total redundancy. The probability of failure of the main and each of the n backup circuits will be equal to [32, 33]: Wj (t) =

n 

ωi (t),

i=1

ω(t) – probability of no-failure operation of the i-th element in the j-th circuit. Probability of failure of this circuit Qi (t) = 1 − Wj (t) =

n  i=1

ωi (t)

O. Romanov et al.

11

12

1i

1n

21

22

2i

2n

j1

j2

ji

jn

m1

m2

mi

mn

m

250

n Fig. 18. Model n of an elemental system with m - multiple total reservation

The probability of failure of the entire system (main and m parallel circuits) is equal to Q(t)

=

m+1 

Qj (t) =

j=1

m+1 

 1−

j=1

n 

 ωi (t)

i=1

Finally, the probability of system uptime W(t)

=

1 − Q (t)

=1−

m+1  j=1

 1−

n 

 ωi (t)

i=1

Assuming that the reservation system is composed of identical elements ωi (t) = ωj (t) = ω, qi (t) = qj (t) = q, rewrite the expressions to determine the probability of uptime m+1  Wgen. = 1 − 1 − ωn then the probability of failure m+1  m+1  = 1 − (1 − q)n Qgen. = 1 − ωn Consider the generic reservation method applied to IMS. For definiteness, we will use a one-time duplication of all the main elements of the multimedia platform. For an IMS system, this would look like a duplication of the entire sequential chain P-CSCF, I-CSCF, S–CSCF. The simplified IMS architecture using the generic reservation method will look like that shown in Fig. 19 [8]. Perform a calculation of the index reliability of the system Fig. 6. This system consists of two parallel-connected functional elements, each of which has «a» index reliability WP−CSCF_I−CSCF_S−CSCF .

Methods for Calculating the Performance Indicators of IMS IMS transport network level

Application level

CSCF1 a

P-CSCF1

I-CSCF1

251

S-CSCF1

b

TAS

CSCF2 c

P-CSCF2

I-CSCF2

S-CSCF2

d

Fig. 19. Simplified IMS architecture using generic reservation

Fig. 20. IMS architecture model using generic reservation

Then the system model of Fig. 19, can be represented as Fig. 20. Then the probability of system uptime WCSCF OB REZ with a generic reservation can be calculated as follows:   WCSCF GEN. RES. = 1 − 1 − WP−CSCFI −CSCFS −CSCF ∗ 1 − WP−CSCFI −CSCFS −CSCF = 1 − (1 − ωP−CSCF ∗ ωI−CSCF ∗ ωS−CSCF )(1 − ωP−CSCF ∗ ωI−CSCF ∗ ωS−CSCF ) The method of separate reservation assumes reservation on separate elements of the system. In general, for a system consisting of n elements with values of reliability indexes (ω1 , ω12 , . . . , ω1i . . . ω1n ), and implying m - multiple reservation [8]. The probability of uptime the main and each of the n backup circuits can be determined from the expression:  m+1 n

1− W (t) = [1 − ω(t)] , i=1

j=1

ω(t) – probability of uptime j-th element in the i-th circuit. Assuming that the reservation system is composed of identical elements ωi (t) = ωj (t) = ω, qi (t) = qj (t) = q, rewrite the expressions to determine the probability of uptime

n Wsep. = 1 − (1 − q)m+1 then the probability of failure

n n  Qsep. = 1 − 1 − (1 − q)m+1 = 1 − 1 − qm+1 Apply the separate reservation method to IMS. For definiteness, we will use a onetime duplication of all the main elements of the multimedia platform. For the IMS system,

252

O. Romanov et al.

it will look like a duplication of individual functional elements: P-CSCF, I-CSCF, S– CSCF. The simplified IMS architecture using the separate reservation method will look like that shown in Fig. 8. We will calculate the index of the reliability of the system Fig. 21. This system consists of three series-connected elements, each of which is duplicated by a backup element. Indexes of reliability of elements are respectively equal: ωP−CSCF , ωI−CSCF , ωS−CSCF .

Fig. 21. IMS split-reservation model

Then the probability of system uptime WCSCF RAZ REZ with split-reservation can be calculated as follows:  WCSCF SEP. RES. = 1 − (1 − ωP−CSCF ) ∗ (1 − ωP−CSCF ) ∗1 − (1 − ωI−CSCF ) ∗ (1 − ωI−CSCF ) ∗ ∗ 1 − (1 − ωS−CSCF ) ∗ (1 − ωS−CSCF ) Let’s make a comparative assessment of the considered methods. In order to evaluate the effectiveness of various reservation methods for the IMS architecture, let us set the initial data on the probability of failure-free operation of functional blocks. Let us assume that all functional blocks of the IMS architecture have the same probability of failure-free performance equal to 0,9. Then, in the absence of a reserve, the reliability of IMS operation will be equal to: WP−CSCF_I−CSCF_S−CSCF = ωP−CSCF ∗ ωI−CSCF ∗ ωS−CSCF = 0, 729. With full global reservation, the IMS’s operational reliability will be: WCSCF GEN. RES. = 1 − (1 − ωP−CSCF ∗ ωI−CSCF ∗ ωS−CSCF )(1 − ωP−CSCF ∗ ωI−CSCF ∗ ωS−CSCF ) = 0, 926559 When using separate reservation IMS operation relia = 1 − − ω ∗ − ω bility will be equal to: W (1 ) (1 CSCF SEP. RES. P−CSCF P−CSCF  ) ∗  1 − (1 − ωI−CSCF ) ∗ (1 − ωI−CSCF ) ∗ 1 − (1 − ωS−CSCF ) ∗ (1 − ωS−CSCF ) = 0, 970 Thus, with the use of full general reservation, the IMS operation reliability can be increased by 1,27 times, and with the use of separate reservation, the IMS operation reliability increases by 1,33 times.

7 Conclusions 1. An analysis of the evolution of the development technologies of modern telecommunication networks shows that the general direction of development can be characterized as follows:

Methods for Calculating the Performance Indicators of IMS

253

NGN (Next Generation Network) - IMS (IP Multimedia Subsystem) - SDN 2. Within the framework of the NGN concept, the following tasks were solved: – creation of a single telecommunication space, ensuring the interaction of networks of various technologies; – distribution of levels at which technology development began to be carried out by specialists in various fields of knowledge independently of each other; – harmonious development of telecommunications, in which networks using efficient technologies began to cover an ever wider market segment, and networks using outdated technologies are gradually dying out; – the rapid introduction of new services without interfering with the functioning of other levels. 3.

4.

5.

6.

7.

8.

The main disadvantage of NGN was the problem associated with the SoftSwitch equipment, which was the main element of the control level. It turned out that the SoftSwitch equipment from different manufacturers is incompatible with each other neither in terms of function, nor in interface. In the process of introducing new modern services in 4G and 4.5G networks, it turned out that a solution to this problem is impossible without bringing the networks of mobile operators in accordance with the requirements of the IMS architecture. The NGN concept has been replaced by the IMS concept, which has eliminated a number of problems. This was achieved thanks to a fundamentally new approach to standardizing the level of management. In IMS, the control layer was standardized in the form of standardization of a set of functions, rather than monolithic physical devices, as it was in NGN networks. Now the management level has become a set of functional blocks, for the practical implementation of which a multiservice platform in the form of a set of proxy servers could be used. From that moment, the problem of ensuring independence from the equipment supplier company was solved. IMS provides a solution to two main tasks. The first task is to ensure the interaction of networks using different alarm systems. The second task is to ensure the interaction of networks at the level of traffic flows. When solving the first problem, all types of signaling are transferred to the use of the SIP protocol. When solving the second problem, all types of transport streams are transferred to the protocol stack of IP networks. These tasks are solved at the control level, the main elements of which are the functional blocks P-CSCF, I-CSCF, S-CSCF. The article provides an analysis of the functioning of the IMS core in solving problems of pairing data streams and signaling streams of networks of various technologies. The list of the main elements involved in the process of solving management problems is determined. This allowed us to develop a number of models for calculating the performance indicators of the IMS management level. To calculate the required throughput indicators, an analysis of traffic servicing processes was carried out using the elements P-CSCF, I-CSCF, S-CSCF. The points of interaction of networks of various technologies with the architecture of IMS are

254

O. Romanov et al.

determined. A model is proposed that allows formalizing the process of servicing incoming traffic from heterogeneous networks and converting it into a standard form in accordance with the requirements of the IMS architecture. Analytical dependencies are obtained, which allow calculating the amount of traffic coming from heterogeneous networks to the elements of the P-CSCF, I-CSCF, S-CSCF control level after its conversion to the standard form of IMS architecture. The obtained results allow us to determine the performance and throughput requirements of the IMS control level elements. 9. The requirements for reliability indicators of the IMS management level are very high. To fulfill these requirements, it is necessary to use methods of reservation of control-level elements. In the work, models have been developed that take into account the essence of the process of reservation of elements of the IMS control level. Analytical expressions are obtained that allow calculating the reliability indicators of the control level when using various backup methods. An example of calculating reliability using the method of general reservation of elements and the method of separate reservation of elements is given. Calculations showed that both methods increase the reliability of the system by 27% or more. At the same time, the reliability of an IMS system using a separate backup method is higher than the reliability of an IMS system using a shared backup method. 10. The direction of further research may be consideration of more complex methods of reservation of system elements, a more complete consideration of the features of IMS and consideration of the impact of system cost indicators. For example, it is advisable to consider these issues, taking into account the costs of implementing a specific method of reserving system elements. However, it should be borne in mind that the theoretical complexity of solving the problem will increase sharply.

References 1. Ilchenko, M.Yu., Kravchuk, S.O.: Advances in the telecommunications 2019/according to scientific editorship, Kyiv, 336 p. (2019). (in Ukrainian) 2. Labovitz, K.: Network traffic during a pandemic. Commun. Bull. (04), 17–20 (2020). (in Russian) 3. Quality Control of Mobile Communication Management Services in a Hybrid Environment. In: Mykhailo, I., Leonid, U., Larysa, G. (eds.) Advances in Information and Communication Technologies. Processing and Control in Information and Communication Systems. Lecture Notes in Electrical Engineering, vol. 560 (2019). Print ISBN 978-3-030-16769-1/ISBN 9783-030-16770-7 (eBook) 4. Romanov, O., Mankivskyi, V.: Optimal traffic distribution based on the sectoral model of loading network elements. In: 2019 IEEE International Scientific-Practical Conference Problems of Infocommunications, Science and Technology (PIC S&T), 8–11 October 2019, Kyiv, Ukraine, 09 April 2020. http://dx.doi.org/10.1109/PICST47496.2019.9061296. Print ISBN: 978-1-7281-4182-4 5. Romanov, O.I., Nesterenko, M.M., Mankivsky, V.B.: Application of the regression model of the coefficient of use of channels for forming the plan of load distribution in the network. Bull. NTUU “KPI”. Radio Eng. Ser. Radio Appar. Constr. (67), 34–42 (2016). (in Ukrainian)

Methods for Calculating the Performance Indicators of IMS

255

6. Romanov, O.I., Oryschuk, M.V., Hordashnyk, Y.S.: Computing of influence of stimulated raman scattering in DWDM telecommunication systems. In: 2016 International Conference Radio Electronics & Info Communications (UkrMiCo) (2016). https://doi.org/10.1109/ukr mico.2016.7739622. eISBN: 978-1-5090-4409-2 7. Globa, L., Skulysh, M., Romanov, O., Nesterenko, M.: Quality control for mobile communication management services in hybrid environment. In: Ilchenko, M., Uryvsky, L., Globa, L. (eds.) Advances in Information and Communication Technologies. UKRMICO 2018. Lecture Notes in Electrical Engineering, vol. 560. Springer, Cham (2019). https://doi.org/10. 1007/978-3-030-16770-7_4. ISBN 978-3-030-16769-1 8. ITU-T Recommendation M.3371 of October 2016: Requirements for service management in cloud-aware telecommunication management system. https://www.itu.int/rec/T-REC-M.337 1/en 9. Lemeshko, O.V.: Fault-tolerance improvement for core and edge of IP network. In: Lemeshko, O.V., Yeremenko, O.S., Tariki, N., Hailan, A.M. (eds.) XIth International Scientific and Technical Conference Computer Science and Information Technology (CSIT 2016). Conference Proceedings – Lviv, pp. 161–164. Polytechnic National University, Lviv, 6–10 September (2016) 10. Romanov, O.I., Nesterenko, M.M, Mankivsky, V.B.: Application of the regression model of the coefficient of use of channels for forming the plan of load distribution in the network. Bull. NTUU “KPI”. Radio Eng. Ser. Radio Appar. Constr. (67), 34–42 (2016). (in Ukrainian) 11. Yeremenko, O., Lemeshko, O., Persikov, A.: Secure routing in reliable networks: proactive and reactive approach. In: Advances in Intelligent Systems and Computing II, CSIT 2017, vol. 689, pp. 631–655. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-70581-1_44 12. Pathak, R.: Is the traditional IMS (IP multimedia subsystem) fitting into 5G world? July 01, 2019. https://www.netmanias.com/en/post/blog/14362/5g/is-the-traditional-ims-ipmultimedia-subsystem-fitting-into-5g-world 13. Ilyas, M., Ahson, S.A., Weik, P., Vingarzan, D., Magedanz, T.: IP Multimedia Subsystem (IMS). https://www.taylorfrancis.com/books/e/9781315219011/chapters/10.1201/978 1315219011-14 14. Ericsson’s “IMS innovation platform” for web real time Communications turns any device with a web connection into an open communications device, Ericssion. http://www.ericsson. com/news/1669129. Accessed 8 Jan 2013 15. Global Market Size of IP Multimedia Subsystems (IMS) 2016–2026: Published by Shanhong Liu. https://www.statista.com/statistics/718069/worldwide-ims-market-size/. Accessed 29 Nov 2018 16. Afanasieva, L., Minochkin, D., Kravchuk, S.: Providing telecommunication services to antarctic stations. In: Proceedings of the 2017 International Conference on Information and Telecommunication Technologies and Radio Electronics,2017, UkrMiCo, pp. 1–4. IEEE Conference Publications, Odessa, Ukraine (2017) 17. Ilchenko, M., Kaidenko, M., Kravchuk, S., Khytrovskyy, V.: Compact troposcatter station for transhorizon communication. In: Proceedings of the 2017 International Conference on Information and Telecommunication Technologies and Radio Electronics (UkrMiCo) 11– 15 September, 2017, pp. 1–4. IEEE Conference Publications, Odessa, IEEE Xplore Digital Library, Ukraine (2017) 18. Umber Iqbal, Y.J.: SIP-based QoS management framework for IMS multimedia. IJCSNS Int. J. Comput. Sci. Netw. Secur. (2010) 19. Romanov, O.I., Nesterenko, M.M., Veres, L.A.: Integration Of modern protocols IP-telephony in IMS architecture. In: 2018 International Conference on Information and Telecommunication Technologies and Radio Electronics (UkrMiCo), 10–14 September, 2018, Kyiv, Ukraine, 26 March (2020). https://doi.org/10.1109/UkrMiCo43733.2018.9047587. (e)ISBN: 978-1-5386-5264-0

256

O. Romanov et al.

20. Romanov, O.I., Hordashnyk, Y.S., Dong, T.T.: Method for calculating the energy loss of a light signal in a telecommunication Li-Fi system. In: Proceedings of the 2017 International Conference on Information and Telecommunication Technologies and Radio Electronics (UkrMiCo), Odessa, Ukraine, https://doi.org/10.1109/ukrmico.2017.8095404. eISBN: 978-1-5386-1056-5 21. Elmostafa, B., Raouyane, B., Belmekki, A., Bellafkih, M.: Secure SIP signaling service in IMS network. In: 2014 Conference - 9th International Conference on Intelligent Systems: Theories and Applications, SITA 2014, Rabat, Morocco, vol. 1. https://doi.org/10.1109/sita. 2014.6847291 22. Romanov, O.I., Fediushyna, D.M., Dong, T.T.: Model and method of Li-Fi network calculation with multipath light signals. In: 2018 International Conference on Information and Telecommunication Technologies and Radio Electronics (UkrMiCo), Kyiv, Ukraine, 10–14 September, 2018, 26 March (2020). https://doi.org/10.1109/ukrmico43733.2018.9047550. (e)ISBN: 978-1-5386-5264-0/ 23. Kravchuk, S., Afanasieva, L.: Wireless cooperative relaying without maintaining a direct connection between the source and target receiver terminals. Inf. Telecommun. Sci. 2, 5–11 (2019) 24. Carella, G., Corici, M., Crosta, P.S.: Cloudified IP multimedia subsystem (IMS) for network function virtualization (NFV)-based architectures. In: 2014 IEEE Symposium on Computers and Communication (ISCC), Proceedings - International Symposium on Computers and Communications (June 2014). https://doi.org/10.1109/ISCC.2014.6912647 25. Romanov, O.I., Nesterenko, M.M., Veres, L.A., Hordashnyk, Y.S.: IMS: model and calculation method of telecommunication network’s capacity. In: Proceedings of the 2017 International Conference on Information and Telecommunication Technologies and Radio Electronics (UkrMiCo), Odessa, Ukraine, https://doi.org/10.1109/ukrmico.2017.8095412. eISBN: 978-1-5386-1056-5 26. Muhammad, A., Muhammad, R., Ulraja, S.: Investigation IMS architecture according to security and QoS context. Chalmers University of Technology, Department of Computer Science and Engineering (Chalmers) (2014). https://hdl.handle.net/20.500.12380/202662/ 27. V10.0.0, E.T.: Security architecture (3GPP TS 33.102 version 10.0.0 Release 10). ETSI TS 133 102 (May 2011). http://www.etsi.org/deliver/etsi_ts/133100_133199/133102/10.00.00_ 60/ts_133102v100000p.pdf 28. GPP Release 7: Provide a unified IMS supporting heterogeneous network access technologies (March 2007) 29. Yu, X., Zhang, X.N.: Research of SIP signaling delay in IP multimedia subsystem. Appl. Mech. Mater. 644, 4387–4390 (2014) 30. Wuthnow, M., Stafford, M., Shih, J.: IMS: a new model for blending applications (Informa Telecoms & Media). Auerbach, vol. 368 (2009) 31. Deart, V.Y.: Multiservice Network. (Softswitch/IMS). Briz-M, Moscow, 198 p. (2011). (in Russian) 32. Atzik, A.A., Goldstein, A.B., Goldstein, B.S.: Megaco/H.248 Protocol: Reference. SPb, BHV Petersburg, 816 p. (2014). (in Russian) 33. Levin, V.I.: Logical theory of reliability of complex systems. Energoatomizdat, 487 p. (1985). (in Russian)

Implementation Biometric Data Security in Remote Authentication Systems via Network Steganography Galyna Liashenko1(B)

and Andrii Astrakhantsev2(B)

1 Faculty of Infocommunications, Department of Information and Network Engineering,

Kharkiv National University of Radio Electronics, Kharkiv, Ukraine [email protected] 2 Institute of Telecommunication System, ITM Department, National Technical University of Ukraine “Igor Sikorskiy Kyiv Polytechnical Institute”, Kyiv, Ukraine [email protected]

Abstract. The need of protection user biometric information increases in consequence of the development of biometric authentication systems. This is due to the fact that fingerprints, iris patterns, face geometry and other biometric data are unique and cannot be replaced. There are various methods for protecting biometric data, but the probability of compromise remains when this data is transmitted over the network. The article presents a method of remote biometric authentication using network steganography for different systems. This improves the reliability of the protection of user biometric data. Analysis of existing methods of network steganography, methods of biometric authentication. Synthesis of new method of remote authentication, that will increase the security of user biometric data from unauthorized access. Modeling a remote authentication system using network steganography methods and user biometric data is the method of this work. The common methods of biometric authentication, existing methods for their protection and existing network steganography methods were analyzed. The method of remote biometric authentication using network steganography for various systems is presented. The remote authentication system using network steganography was simulated. The resistance of the investigated methods to detection was evaluated. The effectiveness of the proposed method for protecting biometric data was investigated. Keywords: Network steganography · Biometric authentication

1 Introduction Biometric technologies are increasingly entering our lives. Things have already become familiar, such as, for example, a fingerprint sensor in a smartphone, biometric data in foreign passports, face recognition in a bank. Users biometric data is used for biometric authentication in payment systems, in smart home systems, and for remote authentication in various applications. Biometric authentication is becoming increasingly popular © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 257–273, 2021. https://doi.org/10.1007/978-3-030-58359-0_14

258

G. Liashenko and A. Astrakhantsev

in banking systems, in healthcare systems, on-line training with the development of information and communication technologies. Biometric authentication applies for authentication in the smart city systems, for secure access to IoT services. The use of biometric data gives the following advantages; the user does not need to remember and enter complex passwords on a mobile device. The password can be cracked or forgotten, and the identification card is what we have and can be stolen. There are static and dynamic methods of biometric authentication. Static methods use physical functions such as fingerprints, iris, retina, hand geometry, facial geometry. Dynamic methods use human behavior - such as signature authentication, keyboard handwriting, voice authentication [1, 2]. Figure 1 shows the classification of biometric authentication methods. In terms of confidentiality, the greatest concern about the use of biometrics has been identified as a result of the preservation and misuse of biometric data. One of the most common applications of biometric authentication is its use in providing remote access to a telecommunications network. The quality of the biometric sample is also important. High-quality data does not suffer from noise problems [3, 4]. Multimodal biometric systems are also used to increase the reliability of the biometric system. Such systems use more than one sensor and obtain multiple user biometric data and use them in certain combinations (e.g., iris and fingerprints). This increases the reliability of the systems.

Fig. 1. Classification of biometric authentication methods

2 Basic Statement of Considered Problem The generalized biometric system consists of modules such as a sensor, a module for extracting biometric features, a module for comparison, as well as a module of the system database that stores biometric templates. At the registration stage, the biometric system records a sample of the user’s biometric characteristics using a sensor - for example, the iris of the eye is scanned, a fingerprint

Implementation Biometric Data Security in Remote Authentication Systems

259

or a face image is taken. Then the required modality is removed from the biometric characteristic, the vector of biometric features is calculated. The system stores the vector in a database along with other identifiers, such as a name or ID number. During the authentication phase, the user provides another biometric sample that is compared to a template on the server or device. Authentication is successful if the new sample matches the template according to some matching function. It returns the degree of correspondence between the template and the query. The system accepts the application only if the compliance rating exceeds a predetermined threshold. Then the user receives a response from the system as to whether further access is granted or denied. Using various technical means, attackers can attack authentication systems to gain unauthorized access to and modify protected data. Therefore, the task of protecting biometric data is very important. The biometric system has several points where biometric data can be compromised. Researchers [5, 6] have identified various types of attacks on such systems. These include attacks such as Trojan, phishing attacks, attacks on the communication channel. Possible attacks on biometric systems are shown in Fig. 2.

Fig. 2. Diagram of biometric authentication with threats

An attacker could violate privacy and alter a user’s biometric information, substitute a fake biometric sensor, intercept legal biometrics, retransmit for access, modify software algorithms that decide to grant access, associate a fake template with an existing user, or register an attacker. Biometric systems can be attacked by exposing a parameter, namely the key used to transform biometric templates. In the case where the transformation is reversed, the original biometrics can be reconstructed. In this case, security is a secret. If the transformation is not reversed, the attacker may try to roughly restore the original biometric patterns. It has also been investigated that when several transformed templates are generated from the same original template, they can be broken by a method called “Attack via Record Multiplicity”. In particular, by giving a transformed template, an attacker can find inverse solutions by reversing the transformation [7]. In recent years, several biometric protection schemes have been proposed that seek to protect the confidentiality of keyless biometric templates. For example, one such scheme is a visual cryptography method that decomposes a biometric image into two noise like images, called sheets, that are stored in two different databases. During authentication, two sheets overlap to create a temporary image for matching. One of the limitations of this method is that it requires two common databases to work together, which may not be practical for some applications. Another method of protecting a biometric fingerprint

260

G. Liashenko and A. Astrakhantsev

combines two fingerprints from two different fingers to create a new pattern. Two request fingerprints are required for authentication. This method offers a two-step matching process for matching two query fingerprints to a combined template. One of the advantages of this method is that when using a combined template, the full characteristics of a single fingerprint will not be violated when the database is stolen. In [8] methods of cancelable biometrics were proposed. Also, the system modules are interconnected by communication channels (both internal to the system and external, for example (communication channel with the data comparison subsystem)). Such external channels are also vulnerable to malicious attacks. The paper proposes the use of network steganography methods to increase the security of information in communication channels.

3 Description of the System Under Study In this work was simulated a system of remote authentication of the iris using network steganography for studying the methods of increasing security. This paper investigates such stages of the remote biometric authentication system model as biometric template formation and bio-hash formation, as well as data transmission by telecommunication network (Fig. 3).

Fig. 3. The biometric authentication system

The formation of a biometric template of the iris of the human eye occurs after obtaining a clear image of the eye with biometric sensors and combines the processing of the image of the eye, applying a filter to highlight important characteristics, generating the iris code. Storing an unprotected original iris code in a database is dangerous because if a code is stolen, a person will not be able to generate another one based on the same iris. Therefore, the application of biometric hashing to the generated iris code has been added to the biometric authentication scheme. Typically, a bio-hash combines a set of random vectors specific to a user with biometric features. After generating a bio-hash based on the generated iris code, the bio-hash is entered into the database. The first time when receiving the iris code and enter it in the database as a reference, it is called user registration. When a user is recognized for authentication, the bio-hash is compared with the bio-hashes stored in the database. The comparison can be made using one of two metrics called Hamming distance and Jacard similarity. This paper investigates several bio-hash algorithms, as well as compares their efficiency with each other and their error rates. The above algorithms are studied in the conditions of generating the iris code according to the algorithm, which was presented in [9].

Implementation Biometric Data Security in Remote Authentication Systems

261

The ideal biometric characteristic should be convenient to use, resistant to fakes, not dependent on the aging of a person, the influence of diseases, and not lead to authentication errors. To study the effectiveness of the software implementation of iris recognition, images from the database CASIA-IrisV3 collected by the Chinese Academy of Sciences’ Institute of Automation (CASIA) were selected [10]. The images available in this database were taken using a special camera with a circular LED array of the near-infrared spectrum with a corresponding light flux to image the iris. Thus, high-quality images from this database are well suited for studying the detailed textural features of iris images. The resolution of such images is 320 × 280 pixels. User identification is based on the characteristics of the iris of the eye, such as furrows, freckles, crypts, and pigment spots. Eye color information is not important for recognition. Therefore, all images should be provided in the grayscale. Because a photo of the eye usually has many clear outlines, they should be “blurred” to clearly distinguish the pupil circle and the circle that is the boundary between the iris and the sclera. Due to this “blur”, the contour lines will be significantly enlarged, which will help highlight larger objects in the image. A Gaussian filter is used for this purpose. Gaussian filters have the property of not overlapping the input of the step function, minimizing the rise and fall time. This behavior is closely related to the fact that the Gaussian filter has the minimum possible group delay. Typically, such a filter is used to reduce the noise level in the image [9]. A Kenny detector is used to highlight the pupil circle. The Kenny detector is an edge detection operator that uses a multi-step algorithm to detect a wide range of edges in images. The algorithm of action of the Kenny detector consists of the following stages: – – – – –

use a Gaussian filter to smooth images and remove noise; finding the gradient of image intensity; use of a double threshold to determine potential edges; tracking the edge with hysteresis; suppression of all other ribs that are weak and not connected to the strong edges.

Kenny algorithm contains several parameters that can be adjusted. They can affect the calculation time and the efficiency of the algorithm. These include the size of the Gaussian filter and threshold values. Among the methods of edge detection developed so far, the Kenny algorithm is one of the most accurate among the identified methods and provides good and reliable detection. Due to its optimality, it meets many criteria for detecting ribs and has a simple implementation process. It is currently one of the most popular edge detection algorithms. The Hough transform is used to define circles. Hough transform is a feature extraction technique used in image analysis, computer vision and digital image processing. The purpose of the transformation is to identify imperfect specimens of objects within a certain class of figures by a voting procedure. The classical Hough transform is most often used to detect regular curves, such as lines, circles, and ellipses. After applying the Hough transform to the image, the coordinates of the center and the radius of the pupil were determined. According to the research, the diameter of the iris is usually between ten and thirteen millimeters. It was determined that for the studied images this

262

G. Liashenko and A. Astrakhantsev

distance corresponds to one hundred pixels. The next stage of image processing is its normalization. The normalization step uses a geometric normalization scheme to convert the selected iris image from a Cartesian coordinate system to a polar one. The result is a rectangular image that is used for further processing. Image normalization has the following advantages. 1) It explains changes in pupil size due to changes in external light that can affect the size of the iris. 2) It ensures that the irises of different faces are not reflected in the overall area of the image, despite variations in pupil size. 3) It allows the registration of the iris during the appropriate stage with a simple translation operation, which can take into account the rotation of the eyes and head in the plane. As mentioned in the third section, the mechanism responsible for the diversity and complexity of the iris code is a filter that “enhances” the characteristics of the iris of the human eye. One of the most effective such mechanisms is the Gabor filter. It is a linear filter used for texture analysis. The Gabor filter analyzes whether there is a certain content of frequency in the image in specific directions in the localized area around the point or area of analysis. Some scientists claim that the Gabor filter is similar to the human visual system. They have been found to be particularly relevant for the presentation of texture. The next step in iris biometric authentication is to generate an iris code. During code generation, the value of each pixel is analyzed, and, depending on this, the value in the code is determined: 0 or 1 [9].

4 Research of Bio-Hash Formation Algorithms Biometric data in case of compromise cannot be changed. To solve this problem, various authors have proposed protection schemes for biometric templates. One of these methods is called canceled biometric methods. Such methods convert user biometric data in accordance with the specified parameters from the user password or key. The original biometric sample of the user is not stored in the database, but only the converted template is stored and authentication is based on the template. In various applications, various templates can be created for the user [11, 12]. Templates transformed in this way must be irreversible, it must be computationally difficult to restore, the original biometric sample, the transformation must be one-sided. Templates must be unique and incoherent, that is, different versions of templates from the same source data should not be comparable. Also, when a template is compromised, it should be possible to recall the template and create a new template, this enhances the security of biometric data. There are various methods for generating a cancelable biometric template. In [13, 14], the authors proposed classifications of biometric sample protection methods. Some types of methods for creating biometric templates are shown in the Fig. 4.

Implementation Biometric Data Security in Remote Authentication Systems

263

Fig. 4. Cancelable biometric techniques

There are methods based on cryptography, transformation, filters, hybrid methods. One of the earliest methods for creating templates with the ability to cancel biometric data is irreversible conversion. In the article [15], the formation of a canceled template is achieved by applying geometric transformation and biological hashing algorithms. The biometric template is first encrypted using geometric transformation algorithms such as Cartesian transformation, polar transformation, and functional transformation. Then, bio-hashing is applied to the tokenized random number specified by the user to the converted pattern to obtain an encrypted pattern. Also there are methods based on random polynomials for creating secure biometric templates; one of such PolyCodes methods was considered in [16]. In [17], a method was considered in which a randomly generated shuffle key is used to shuffle biometric user data. Mixed biometric data represents an undoable pattern. Cancelable Biometric represents a template protection mechanism where the original biometric sample is intentionally distorted for registration in the identification system. When a canceled biometric scheme is used, a distorted version of the template is retained instead of the original biometric template. Canceled biometrics solves such privacy issues because it prevents the system from retaining the user’s original biometrics, and it is computationally difficult to recover input biometrics. A revocable biometric template is a solution to this problem that can be republished in the event of a hack. Methods of biometric template protection can be divided into inverse and irreversible transformations. The first methods include such methods as BIN-SALT and GRAYSALT, the second - BIN-COMBO and GRAY-COMBO, which are considered in [8]. At the end of the biometric template formation phase, an iris code of 17600 bits was obtained. Then it was converted to an integer type. As a result, the number of elements has been reduced to 560. After this step, it is possible to save the generated iris code to the base, but it is not secure. Accordingly, the means of canceled biometrics were used to protect the biometric code. In cancelable biometric, independent auxiliary data, such as a user-defined password or token, are combined with biometric data to provide a distorted version of the biometric template. In GRAY-SALT, a randomly generated vector was added to or multiplied by an iris pattern. In BIN-SALT, the XOR operation was used for the iris code, and the randomly generated binary key. For both GRAY-SALT and BIN-SALT, information about the iris pattern is hidden by auxiliary data. In the studied program, biometric salinization of the iris code was implemented by the BIN-SALT method. A binary key was generated to the iris code using a random number generator. The length of the key was equal to the length of the iris code mentioned above. The binary key was generated only once and is

264

G. Liashenko and A. Astrakhantsev

common to all biometric templates. The program provides for the generation of a new key in case of theft of the bio-hash of one of the users. It should be noted that there are other schemes of cancelable biometric. In contrast to the described option, a modified user password can also be used as a key. Such a scheme can be performed in two-factor authentication systems. The disadvantage of all variations of biometric salting methods is that in the case of obtaining a binary key or a secret vector, an attacker can easily recover the user’s iris code. The methods of irreversible transformation are deprived of the disadvantage of biometric salinization. Irreversible transformation is conceptually attractive for template protection schemes. In the irreversible transformation, the unilateral transformation function is used to transform the iris pattern. Thus, in the case of abduction, it is impossible to restore the original code of the iris. Two methods were considered for the irreversible conversion of the iris code: GRAYCOMBO and BIN-COMBO. In GRAY-COMBO, elements of the iris code are rearranged using a random combination, after which operations of addition or multiplication are performed for two randomly selected lines. In BIN-COMBO, the same procedure is performed for the iris code, but with an XOR or XNOR operation. Thus, the original data of the iris code was distorted, changed by operations of addition or multiplication between two randomly selected lines. Thus, the criterion of irreversibility is fulfilled [8]. The method of irreversible BIN-COMBO transformation was implemented in the studied program. This method was implemented as follows. Using a random number generator, a sequence was created in which the number of elements was equal to the number of lines in the original iris code. Each value of the generated sequence determines the number of elements on which to wrap the elements of the line. After that, one prime number was selected using a random number generator. All lines whose sequence numbers were divided by this number were connected to the following lines by an XOR operation. The disadvantage of this method is that due to the reduction of the iris code after modifications, the efficiency of the recognition system has deteriorated. Another method for changing the iris code is Min-Hashing. During the operation of the Min-Hashing algorithm, the index of the first-bit encounter, which is equal to 1, is written for a number of binary vectors whose rows have been shuffled. Let A and B be two index vectors generated from a binary vector, and let h be a hash function that calculates hashes for the elements of these sets. Next, is necessary to define the function hmin (S), which computes the function h for all members of any set S and returns its smallest value. Then you need to calculate hmin (A) and hmin (B). Comparing the values of hmin (A) and hmin (B) will not work, because the probability that they will be one hundred percent equal is very low. This problem can be solved using the Jacquard similarity factor. The coefficient measures the similarity between sets. According to the described Min-Hashing algorithm, a software implementation of the third specified method of protection of biometric templates was made. Biometric templates, that have to identify one person and taken by the sensor at different times are not completely identical, as they can be made at different angles, lighting, or other environmental factors. Byte code made from the template will completely change the hash function. This can lead to recognition errors, even though the changes were minor. Special methods should be used to solve this problem. Biometric

Implementation Biometric Data Security in Remote Authentication Systems

265

template protection methods retain a modified version of the biometric template and disclose as little information as possible about the original biometric feature without losing the ability to identify the person. Such methods must meet the following criteria: – cross-matching of protected templates must be impossible; – it should be possible to cancel the compromised template and create a new one from the same biometric data; – the efficiency of the biometric system should not be impaired by the template protection system. Biometric hashing is one of the transformation-based methods in which the user’s biometric template is converted to a secure binary string. One advantage of biometric hashing is the ease of recalling the transformed template by changing the corresponding secret key. In addition, using the same biometric data, the user can be recognized for different services using different biometric hashes generated from different secret keys. Thus, two records represented by two different systems cannot be linked, and the user’s activity remains private. The efficiency and quality of the software implementation of the above algorithms were tested by a number of subsequent studies. It was also assessed how much the use of biometric template protection methods affects the recognition efficiency. For all biometric template protection methods and for the unprotected iris code, a threshold value was chosen for the Hemming distance or for the Jacquard’s similarity factor, the comparison with which determines whether to grant access to the user or reject it. This parameter was determined experimentally. To do this, the value of Hemming’s distance or Jacquard’s similarity coefficient was calculated by comparing each biometric template with each. Based on the obtained experimental data, it is possible to distinguish the smallest and the largest distances of Hemming. After using the BIN-SALT method to protect the biometric template, the smallest value of the Hemming distance increased and became equal to 58%, and the largest value decreased and began to be equal to 78%. Thus, the Hemming distance threshold in the case of biometric salinization was chosen to be 78%. The lowest value for the protected iris code using the BIN-COMBO method is 58%, and the highest is 76%. Due to this threshold, 77% was chosen. To study the effectiveness and stability of the above methods of protection of biometric templates in combination with the same methods of image processing specified in the second section, Perlin noise was applied to the images. This noise is based on a function that has a pseudo-random appearance, but all its visual details are the same size. This property makes it easy to control the noise. You can insert multiple large-scale copies of Perlin noise into mathematical expressions to create a wide variety of procedural textures. To superimpose Perlin noise on the image of the iris, a software simulation of the Perlin noise overlay process was developed using the Java programming language. This application involves the application of noise to the image at a given percentage. The selected noise was applied to each eye image at a different percentage. Then, the percentage of noise corresponding to the threshold value of Heming’s distance or Jacquard’s similarity coefficient for each of the studied methods of protection of the biometric template and the original code of the iris was determined experimentally.

266

G. Liashenko and A. Astrakhantsev

threshold percentage of noise at unprotected iris code is from 0.21% to 1.5% at the threshold value of Hemming’s distance of 80%, threshold percentage of noise at the use of BIN-SALT from 0.17% to 2.8% at the threshold value of the distance of Hemming of 78%, threshold percentage of noise at the use of -COMBO from 0.18% to 2.15% at a threshold value of Hemming’s distance of 77%, threshold percentage of noise when using Min-Hashing from 0.17% to 3.7% at a Jacquard factor of 0.14. To analyze the noise level introduced by the noise threshold, a signal-to-noise ratio (SNR) value was determined for each image using different methods of protecting the biometric template and their absence. The signal-to-noise ratio is also used to determine the image quality characteristic [18]. The sensitivity of a digital imaging system is usually described in terms of the signal level that gives the SNR threshold level. It is possible to estimate the probability of occurrence of errors at recognition for each algorithm by means of the following indicators. 1) False acceptance rate (FAR). This indicator is a percentage threshold that determines the probability that one person can be mistaken for another. 2) False rejection rate (FRR). This indicator represents the probability that the registered person may not be recognized by the system. As the number of erroneous omissions decreases, the number of erroneous deviations will increase, and vice versa. The point where the lines intersect is also called: equal error rate (EER). At this point, the percentage of erroneous acceptances and erroneous deviations is the same [19]. Usually, lowering the FAR to the lowest possible level will lead to a sharp increase in FRR. Therefore, the more secure the access control in the biometric system, the less convenient it will be for users, as they are much more likely to be mistakenly rejected by the system. Typically, the FAR and FRR can be configured in the security software by adjusting the appropriate criteria. In the human iris recognition program understudy, this can be done by adjusting the tolerance threshold, which was determined experimentally above for each of the biometric pattern protection methods and the unprotected iris code. Calculated from FAR and FRR for the simulated algorithms are shown in Fig. 5.

Fig. 5. Values of the FAR and FRR for the studied methods.

Implementation Biometric Data Security in Remote Authentication Systems

267

Among the algorithms with biometric template protection, the algorithm using BINCOMBO to create a bio-hash has the closest to the ideal indicators. Instead, the best FRR of all algorithms is the Min-Hashing algorithm to create a bio-hash. This determines the highest value of the FAR indicator for this algorithm. Based on this, we can conclude that the system using Min-Hashing will be quite user-friendly. The FAR values among the algorithms using biometric template methods are slightly different, and for MinHashing and BIN-COMBO are equal. At the same time, the FRR is much better for the algorithm using Min-Hashing to create a bio-hash. Thus, it can be concluded that to provide a secure biometric system with user-friendly use, it is better to use an algorithm using Min-Hashing to create a bio-hash.

5 Application of Network Steganography To protect user biometric data from compromise, a large number of different schemes are presented. Most research uses cryptography to protect user data. For example, in [20], an iris template protection scheme was proposed that combines canceled biometric data with FCS to protect long cryptographic keys without compromising recognition accuracy. In [21] a cryptographic method of fuzzy storage was proposed for the development of a reliable fingerprint verification system. A new biometric cryptosystem for vector biometry called symmetric key encryption based on the Rivest key model was proposed in [22]. Steganography is the science of the hidden transmission of information by keeping secret the fact of transmission. Cryptography protects the content of the message, and steganography protects the very fact of the presence of any hidden messages. There are various methods of steganography - information can be hidden in images, video, audio. It is also possible to hide information in the fields of data transfer protocols [23]. The purpose of this work is the analysis of existing methods of network steganography, methods of biometric authentication. Synthesis of a new method of remote authentication that will increase the security of user biometric data from unauthorized access. And also the research of efficiency of methods of network steganography in case of carrying out remote authentication by an assessment of concealment of the specified methods and their resistance to noise in communication channels. An effective covert channel must not be detected by an attacker and must ensure a high degree of confidentiality. The paper simulates the transfer of a biometric template by the method considered in [24]. This method of network steganography hides data in TCP segments. In the TCP segment, you can embed data in fields such as Window Size, TCP Option, Acknowledge (ACK) number. In this paper, the Window Size field was used to hide data in the TCP segment, the order of modification of which was considered in [25]. Its content was to enter the data hidden in this field, after which the TCP segment was transmitted to the receiver side, and already on its side was the recovery of hidden data. The second method implements data hiding in HTTP headers hereinafter referred to as the NS-HTTP method. In this method, different characteristics of HTTP messages can be used for covert data transmission. These include modifications of the order of

268

G. Liashenko and A. Astrakhantsev

the titles, their structure, and content [26, 27]. The software model that implements the NS-HTTP method corresponds to the client-server architecture: on the client-side, the message is converted to binary format and encoded as spaces within the HTTP request headers. Zero is encoded as space, and one is a double space. The first place after the colon in HTTP headers is not used to hide data. In the case of double space, it is visually very noticeable. Also, to provide more bandwidth, space are added just before the end of the header, which is harder to notice. The third method studied is the method of network steganography based on hiding data in ICMP headers (NS-ICMP method). To implement this method, a program was also developed that has two layers, one of which is responsible for the graphical interface, cryptography and data compression, and the second - for embedding information in ICMP headers. By the standard, ICMP-packet has a format, which is presented in Fig. 6.

Fig. 6. ICMP packet header structure

Both layers are separate programs that interact with each other through Unix interprocess communication (IPC). The program uses compression and encryption of the information that needs to be hidden, data was encrypted using the Advanced Encryption Standard (AES) 265. The developed program supports two modes of data embedding within the NS-ICMP method. The first mode - “Fast”, it allows you to hide 60 bytes of information in one stegocontainer. This mode uses fields such as Identifier, Sequence number, and the data field. The second mode - “Safe”, it changes only two fields: Identifier and Sequence number. Therefore, the second mode has a lower bandwidth compared to the “Fast” mode. For two modes, the 64-byte packet was selected. As these methods are to be used for remote biometric authentication, noise immunity in open communication channels is of great importance when evaluating their effectiveness.

6 Investigation of Resistance to Detection and Noise of Implemented Methods of Network Steganography To study the resistance to the detection of implemented methods of network steganography, an analysis of the impact of the transmission of the built-in hidden message on the characteristics of traffic in general. Traffic was intercepted and analyzed using Wireshark. In Fig. 7 shows a histogram showing the dynamics of changes in traffic characteristics using the method of NS-HTTP. You can see that when a small amount of hidden data (24 bytes) is transmitted, the statistical characteristics of the traffic are almost unchanged,

Implementation Biometric Data Security in Remote Authentication Systems

269

but when more data is transmitted, the amount of TCP and HTTP traffic increases significantly. Each new HTTP header is sent, a new connection is established, each time a TCP session is started, which increases the number of TCP segments.

Fig. 7. Study of statistical characteristics of traffic using the MS-HTTP method

In Fig. 8 presents the results of the analysis of traffic characteristics when using the NS-TCP method. As expected, with a small amount of hidden data transmitted, the amount of TCP traffic increased slightly. In turn, as the amount of hidden data (500 kB) increases, the amount of TCP traffic almost doubles, which makes this method of network steganography quite noticeable.

Fig. 8. Study of statistical characteristics of traffic using the MS-TCP method

The NS-ICMP method was studied under the condition of embedding information in two modes. Figure 9 presents the statistical characteristics of the traffic, which were studied for the modes “Safe” and “Fast”, respectively.

270

G. Liashenko and A. Astrakhantsev

Fig. 9. Retrieval of statistical characteristics of traffic using the NS-ICMP method

The results show that the number of ICMP packets has increased significantly when transmitting 500 kB of data in the “Secure” mode, which cannot be said about the “Fast” mode. When using the “Fast” mode, the number of new ICMP packets is much smaller, which indicates that using this method to create a hidden data channel, the possibility of its detection is much smaller. The stability of the methods against the background of additive white Gaussian Noise (AWGN) in the communication channel was also studied. Random nature of noise in the time domain in some cases random nature of noise in the time domain in some cases caused the transmission of a symbol that was distorted in such a way that the receiver interpreted it as another symbol. If errors were made in the transmitted data, the integrity of the system could be compromised. The Bit Error Ratio (BER) was used to evaluate the efficiency of the system. Before transmission by the communication channel, the packets were encoded using the linear coding algorithm 2 Binary 1 Quandary (2B1Q), which is one of the implementations of the amplitude-pulse modulation algorithm with four levels of output voltage without returning to zero levels (Non Return To Zero, NRZ) [28]. Figure 10 shows the results of calculating the bit error rate for one container.

Fig. 10. The results of BER calculation for the considered methods of steganography

Implementation Biometric Data Security in Remote Authentication Systems

271

NS-ICMP methods were less resistant to noise, NS-HTTP methods at dispersion values less than 0.3 hidden data were restored unchanged. However, as the variance value increased, the message was not restored. In the NS-ICMP method, the message was not restored even at a variance value greater than 0.3, and in the NS-TCP method, the data can be restored with slight distortion at a variance value of 0.4. As a result of this comparison of methods, we can conclude that the most noise-resistant method was network steganography, which hides the data in TCP-headers.

7 Conclusions The unique biometric characteristics of a person cannot be changed, so the task of protecting this data from compromise is very important. There are various ways to protect biometric data. They are not transmitted over the network without prior protection, the databases store protected templates instead of the original data. Biometric cryptosystems and biometric systems with the possibility of cancellation are new technologies for the protection of biometric templates to address these issues and increase public confidence and acceptance of biometrics. In recent years, a significant number of approaches to both technologies have been published, modern approaches are considered based on which in-depth discussion and prospects for the future are given. The existing schemes of cancellable biometrics were considered in the work. The work of three methods of biometric template formation was also simulated. The performance indicators of these algorithms were analyzed. For additional data protection during network transmission, it was proposed to use existing methods of network steganography. Also in this work, the software implementation and analysis of network steganography methods based on data hiding, which are transmitted in TCP, HTTP, and ICMP headers against the background of transmission of hidden biometric data, was carried out. The results of the study showed that the worst method was NS-HTTP, which had low resistance to noise because when the value of the variance is less than 0.3, the hidden message was not restored. Also, in the case of using the methods of NS-HTTP and NS-TCP traffic increased sharply several times. In turn, the NS-TCP method is most effective when working with noise communication channels: it allows you to recover a hidden message with little distortion at a variance of 0.4. However, this method of network steganography significantly loses to the “Fast” mode of the NS-ICMP method by the criterion of concealment. Given the specifics of the application of the considered methods of network steganography for remote authentication by a set of criteria for noise/latency, it is recommended to use the “Fast” method of the NS-ICMP method for use in biometric authentication systems. The obtained results show that the use of network steganography increases the security of biometric data. This method can be used to increase data security when authorizing users on different systems.

References 1. Bharadwaj, S., Vatsa, M., Singh, R.: Biometric quality: a review of fingerprint, iris, and face. EURASIP J. Image Video Proc. 1(2014), 34 (2014)

272

G. Liashenko and A. Astrakhantsev

2. Ruban, I., Martovytskyi, V., Kovalenko, A., Lukova-Chuiko, N.: Identification in informative systems on the basis of users’ behaviour. In: IEEE 8th International Conference on Advanced Optoelectronics and Lasers (CAOL), pp. 574–577 (2019) 3. Liashenko, G., Astrakhantsev A.: Investigation of the influence of image quality on the work of biometric authentication methods. In: IEEE International Scientific-Practical Conference Problems of Infocommunications, Science and Technology (PIC S&T), pp. 812–817 (2019) 4. Olesia, B., Iana, M., Nataliia, Y., Oleksii, L., Danyil, T.: System of individual multidimensional biometric authentication. Int. J. Emerg. Trends Eng. Res. 7, 812–817 (2019) 5. Busch, C.: Standards for biometric presentation attack detection. In: Marcel, S., Nixon, M., Fierrez, J., Evans, N. (eds.) Handbook of Biometric Anti-Spoofing. Advances in Computer Vision and Pattern Recognition. Springer, Cham (2019) 6. Mehmood, R., Selwal, A.: Fingerprint biometric template security schemes: attacks and countermeasures. In: Singh, P., Kar, A., Singh, Y., Kolekar, M., Tanwar, S. (eds.) Proceedings of ICRIC 2019. Lecture Notes in Electrical Engineering, vol. 597. Springer, Cham (2020) 7. Topcu, B., Karabat, C., Azadmanesh, M., et al.: Practical security and privacy attacks against biometric hashing using sparse recovery. EURASIP J. Adv. Signal Process. 2016, 100 (2016) 8. Zuo, J., Ratha, N., Connell, J.: Cancelable iris biometric. In: 19th International Conference on Pattern Recognition 2008, pp. 1–4, December 2008 9. Chernikova, V., Astrakhantsev, A., Liashenko, G.: Investigation of characteristics of the biometric identification system based on iris code. Syst. Arms Mil. Equip. 1(53), 195–202 (2018). (in Ukrainian) 10. CASIA Iris Image Database. https://biometrics.idealtest.org 11. Belguechi, R., Le-Goff, T., Cherrier, E., Rosenberger, C.: Study of the robustness of a cancelable biometric system. In: 2011 Conference on Network and Information Systems Security (2011) 12. Ramesh, P., Subbiah, G.: Can cancellable biometrics preserve privacy? Biometric Technol. Today 2017, 8–11 (2018) 13. Rawat, M., Kumar, N.: Cancelable Biometrics: a comprehensive survey. Artif. Intell. Rev. 53, 3403–3446 (2019) 14. Álvarez, F., Hernandez Encinas, L., Sánchez Ávila, C.: Biometric fuzzy extractor scheme for iris templates. In: 2009 IEEE 3rd International Conference on Biometrics: Theory, Applications, and Systems (2009) 15. Parkavi, R., Chandeesh Babu, K.R., Neelambika, T., Shilpa, P.: Cancelable biometrics using geometric transformations and bio hashing. In: Hemanth, D., Smys, S. (eds.) Computational Vision and Bio Inspired Computing. Lecture Notes in Computational Vision and Biomechanics, vol. 28. Springer, Cham (2018) 16. Kaur, H., Khanna, P.: PolyCodes: generating cancelable biometric features using polynomial transformation. Multimedia Tools Appl. 1–24 (2020) 17. Kanade, S.G., Petrovska-Delacrétaz, D., Dorizzi, B.: Cancelable biometrics for better security and privacy in biometric systems. In: Abraham, A., Mauri, J.L., Buford, J.F., Suzuki, J., Thampi, S.M. (eds.) ACC 2011: 1st International Conference on Advances in Computing and Communications, pp. 20–34 (2011). 18. Konakhovich, G.F., Puzirenko, A.Yu.: Czifrova steganografi‘ya. MK-Press (2006). (in Russian) 19. FAR and FRR: security level versus user convenience. https://www.recogtech.com/en/kno wledge-base/security-level-versus-user-convenience. Accessed 20 May 2020 20. Ouda, O., Tsumura, N., Nakaguchi, T.: Effective combination of iris-based cancelable biometrics and biometric cryptosystems. Int. J. Adv. Comput. Sci. Appl. (IJACSA) 10(11), 658–668 (2019)

Implementation Biometric Data Security in Remote Authentication Systems

273

21. Deshmukh, R., Sapkal, S., Karungaru, S., Banerjee, S.: Biometric cryptosystem for fingerprint verification. In: 2nd International Conference on Cognitive Knowledge Engineering, pp. 240– 244 (2018) 22. Lai, Y., Hwang, J., Jin, Z., Kim, S., Cho, S., Teoh, A.: A Symmetric Keyring Encryption Scheme for Biometric Cryptosystems. Inf. Sci. 502, 492–509 (2018) 23. Shostak, N.V., Astrakhantsev, A.A., Romanko, S.V.: Doslidzhennia stiikosti alhorytmiv zakhystu avtorskykh prav na videoproduktsiiu [Investigation of stability of algorithms of copyright protection in video production]. Inf. Process. Syst. 2(148), 138–143 (2017). (in Ukrainian) 24. Dhokate, D., Ghorpade, V.R., Patil, B.S., Patil, S.B.: Steganography in TCP & IP headers. In: 2010 3rd International Conference on Advanced Computer Theory and Engineering (ICACTE), vol. 4 (2010) 25. Giffin J., Greenstadt R., Litwack P., Tibbetts R.: Covert messaging through TCP timestamps. In: Dingledine, R., Syverson, P. (eds) Privacy Enhancing Technologies (PET): Proceedings of the 2nd international conference on Privacy enhancing technologies. San Francisco, CA, USA, 14–15 April 2002, pp. 194–208 (2002) 26. Blasco, J., Hernandez-Castro, J.C., de Fuentes, J.M., Ramos, B.: A Framework for avoiding steganography usage over HTTP. Netw. Comput. Appl. 35(1), 491–501 (2012) 27. Mazurczyk, W., Wendzel, S., Zander, S., Houmansadr, A., Szczypiorski, K.: Information Hiding in Communication Networks: Fundamentals, Mechanisms, and Applications. IEEE Series on Information and Communication Networks Security, 1st edn., p. 256. Wiley, New York (2016) 28. Sherbak, A., Astrakhantsev, A., Sherbak, O., Liashenko, G.: Analiz skritnosti ta stijkosti do shumu v kanalakh zvyazku metodiv merezhnoyi steganografiyi. Probl. Telecommun. 2(23), 89–98 (2018). (in Ukrainian)

Wireless Systems with New Cooperative Relaying Algorithm Liana Afanasieva(B)

and Sergey Kravchuk

Department of Telecommunications, National Technical University of Ukraine “I. Sikorsky KPI”, Kiev, Ukraine [email protected], [email protected]

Abstract. The possibility of using a mobile terminal as a relay node allows another mobile terminal to transmit or receive data on two independent communication lines, realizing additional diversity of signals both in space and in time. This signal transmission procedure is known as cooperative relay technology. 3GPP group, whose applying is being researched and developed in a number of scientific papers, incorporates cooperative relaying technology in the standard 5G. This technology is used to combat signal degradation on radio link, especially in a multi-user network environment, allows to reduce the effect of interference at the edges of the cell, and to improve radio transmission parameters, thereby ensuring the necessary quality of service QoS. Due to growing the number of wireless network devices in transmission area between the sender and the target receiver, there are more potential relay nodes. Therefore, the task of choosing the one for implementing cooperative transmission in practice becomes an important task. The paper presents a new method for selecting the best relay node, which takes into account a number of criteria, namely: node position, bit error rate, energy consumption. Keywords: Cooperative relaying · Relay selection · Wireless communication network

1 Introduction The evolving LTE-A standard is one of the most promising standards for next-generation wireless systems. Next-generation wireless networks are designed to support high data rates, more coverage, as well as lower power consumption and efficient use of available bandwidth. At the same time, mobile terminals should be simple, cheap and small in size. Providing high bandwidth capacity at the edges of the cell is one of the main objectives of the LTE-A standard. The received signal degrades in a wireless environment due to path loss and shadowing from various obstacles. In addition, the signal quality gets worse due to attenuation due to structural interference from multipath components [1, 2]. The cooperative relaying method is used to combat the degradation of signal quality on a radio link, especially in a multi-user network environment. The applying of cooperative relaying methods can improve reliability, increase throughput, improve the use of © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 274–288, 2021. https://doi.org/10.1007/978-3-030-58359-0_15

Wireless Systems with New Cooperative Relaying Algorithm

275

power, and provide adaptability of next-generation wireless networks. The main idea of this method is to use a relay node to create an alternative channel for transmitting duplicate information. The message sent by the source arrives at the target receiver (D) via spatially different channels: first - directly from the source node, second one - through the relay node. Then, the target receiver node combines the two received signal to improve decoding of the message. Thus, cooperative relaying is a joint transmission of messages from the source to the target receiver using relay nodes, in order to improve the quality of service (speed increase, high levels of interference resistance) of the end user. The use of such a relay communication channel can improve network bandwidth due to the implementation of spatial diversity, since at the same time the same fading effects do not affect the channels [3–7]. In addition to spatial separation, transmission via relay nodes provides an effective way to combat radio path losses, and can also improve the energy efficiency of wireless communications. Thus, we can distinguish several scenarios for the use of cooperative relay in LTE networks, which can improve the quality of QoS service:reducing the effect of interference due to spatial diversity,-increasing throughput through resource aggregation, seamless provision of services capacity at the edges of the cell [8–12]. Ensuring the required level of QoS plays an important role in next-generation wireless networks. This is a complex and difficult task to satisfy the various QoS requirements for the services offered. Thus, when using cooperative relay in the LTE\LTE-A network, it is necessary to consider what services are provided to subscribers in the network and what requirements are presented to the data delivery procedure. In the LTE network, subscribers can take advantage of many services: making a voice call, video call, transferring data from the Internet; broadcast video streaming; data transfer online video games, applications, etc. According to Table 1 of the 3GPP TS 123 203 V12.6.0 [13] specification, the following basic transmission requirements must be taken into account under QoS conditions: – requirements for the time delay, ranging from Real Time Gaming, V2X messages with a delay of 50 ms, to a Video (Buffered Streaming) TCP-based service with a delay of 300 ms; – the relative value of packet error loss rate varies in the range from 10−2 to 10−6 depending on the service; – requirements for the guarantee of providing a given bit rate(GBR) or data transfer without guaranteeing the preservation of speed (non-GBR). Due to growing the number of wireless network devices in transmission area between the sender and the target receiver, there are more potential relay nodes. So the task of choosing the one for implementing cooperative transmission in practice becomes an important task. Devices located between the source/sender and the potential target receiver/destination should coordinate the nodes for relaying according to the best conditions for forming the transmission signals. Therefore, the aim of this work is to develop a new method for selecting a relay node for cooperative relaying, which takes into account a set of criteria such as position of the relay node, bit error rate, power consumption at the relay nodes), which reduces the path loss and the shading effect.

276

L. Afanasieva and S. Kravchuk Table 1. Standardized QCI characteristics

QoS Class Identifier (QCI) according to various types of traffic in LTE

Resource type

Priority

Packet delay budget

Packet error loss rate

Example services

1

GBR

2

100 ms

10−2

Conversational voice

2

GBR

4

150 ms

10−3

Conversational video (live streaming)

3

GBR

3

50 ms

10−3

Real time gaming, V2X messages

4

GBR

5

300 ms

10−6

Non-conversational video (buffered streaming)

5

non-GBR

1

100 ms

10−6

IMS signalling Video (buffered streaming) TCP-based (for example, www, email, chat, ftp, p2p and the like)

6

non-GBR

6

300 ms

10−6

7

non-GBR

7

100 ms

10−3

Voice, video (live streaming), interactive gaming

8

non-GBR

8

300 ms

10−6

Video (buffered streaming) TCP-based (for example, www, email, chat, ftp, p2p and the like)

9

non-GBR

9

300 ms

10−6

Video (buffered streaming) TCP-based (for example, www, email, chat, ftp, p2p and the like). Typically used as default bearer

Wireless Systems with New Cooperative Relaying Algorithm

277

2 Overview of Relay Node In LTE networks, relay nodes are energy-efficient eNodeBs that provide improved coverage and capacity at the edges of the cell. One of the main advantages of the relay is to expand LTE coverage in targeted areas at low cost. Figure 1 shows an example deployment of a relay node in an LTE-A cellular system. Note that the relay nodes (RN) can be placed at the edge of the cell to increase the transmission distance of the cell. Deploying RS near the edge of the cell will help increase throughput or, conversely, expand the coverage area of the cell [14–18]. In the LTE 3GPP standard, there are two types of relays - transparent and nontransparent [14]. Type I or non-transparent relay node allows a mobile station located far from the base station to access the base station. For this purpose, the relay node transmits a common reference signal and control information for the base station. The main function of the non-transpare relay node is to increase the signal propagation range, as well as expand the service area (see Fig. 1a). Thus, by providing communication and data services for remote mobile stations, Type I relay nodes can improve system throughput. (a)

(b)

Fig. 1. The relay node in an LTE-A cellular system. a Type I or non-transparent relay node; b Type II or transparent relay node

278

L. Afanasieva and S. Kravchuk

Type II or transparent relay node has a direct link to the base station. It allows improving the quality of service and throughput of the communication link between the base station and the mobile station, which can be located both in the coverage area of the base station and beyond (see Fig. 1b). The type II relay node does not transmit a common reference signal or control information; its main function is to increase the system capacity due to cooperative diversity and transmission power amplification for local mobile stations. As shown in Fig. 1a for transparent relaying nodes during cooperative relaying, one of the main parameters of transmission efficiency is the territorial location in the cell relative to base station eNodeB and mobile station (MS). Depending on the location of the RN relative to the base station and the mobile station, the appropriateness (efficiency) of the application of cooperative transmission is determined. Figure 2 shows the layout of the relay node, where d1 , d2 , d is the distance from eNodeB to RN, from RN to RN MS and from eNodeB to MS, respectively. Where, d1 + d2 ≥ d , and d2 < d . Depending on the attenuation of the radio signal during its propagation, we consider the effective ratio of these quantities.

Fig. 2. Cooperative relaying model

When the relay node is located close to one node, the distance to the other node will be large, which leads to strong attenuation of the signal on the receiving side. In urban environments in addition to attenuation on the propagation path, it is also possible that obstacles will appear in the signal propagation path, which will lead to the manifestation of the multipath effect. When a co-operator is located nearby with one node, the distance to the other node will be large, which leads to strong attenuation of the signal on the receiving side. In an urban environment should be taken into account not only the path attenuation, but also the possibility of obstacles in the path of the signal, which leads to the manifestation of the effect of multipath. When the cooperation node is located nearby, for example, with eNodeB, RN is in the line of sight, in which case the signal attenuation will be small. On the other hand, a large distance d2 to MS can lead to the manifestation of the multipath effect and strong

Wireless Systems with New Cooperative Relaying Algorithm

279

attenuation of the signal at the receiving side. In reverse ratio values d1 and d2 , it is likely that the relay node itself will take the signal from the eNodeB to attenuation and distortion (see Fig. 3).

Fig. 3. The dependence of the change in signal attenuations on the distance between the transmitting and receiving terminals

When the cooperative relay scheme is activated (Fig. 1), an increase in SNR is observed. From the SNR analysis, it is seen that the cooperative relay mechanism allows increasing the SNR value on the MS, and the greater the deterioration (attenuation) on the line of sight between the eNodeB and the MS, the greater the gain from the inclusion of the cooperation scheme. In both cases, the retransmission signal with errors deteriorates the interference situation at the receiving side and, therefore, cooperative relaying will not be effective. The best mode is when the relay node is located in the second half of the cell at a distance of 0.5–0.8 from the total distance d of the eNodeB - MS channel (see Fig. 4). In [33] it has also been shown that for a given value of bit-error probability of 10−3 , when the relay node is placed at a distance of about 0.7 from the total distance, an energy gain of up to 4 dB is achieved.

280

L. Afanasieva and S. Kravchuk

Fig. 4. The dependence of the signal-to-noise ratio on the distance between the transmitting and receiving terminals

3 Related Work In the general, the transmission of information using cooperative relaying can be divided into three main stages: direct transmission, relay selection, cooperative transmission. During the direct transmission, the source transmits its data to the target receiver, and the relay RN (or potential relays) also try to receive it (“overhear the transmission”). At the stage of cooperative transmission, the relay R transmits data to the target receiver, but only if the D could not receive data from the source during the direct transmission. The following factors influence any relay selection algorithm: the number of relays (single or multiple choice) and the mechanism of selecting ones (centralized or distributed). First of all, it is necessary to determine whether a single or multiple relays will be selected for cooperation. To improve the quality of service QoS in [20–22] uses a lot of repeaters. To improve the quality of service QoS in [20–22] proposed to use multiple relays for transmission. However, the range of interference will increase in proportion to the number of participating in the transmission relays, and may exclude the possibility of reuse of spatial frequency in wireless networks [23]. Thus the need to minimize the number of relay nodes so as to reduce the interference range and to satisfy the required QoS constraints [13]. When co-operation is used to improve channel reliability, having a few relay nodes over a long distance can be more effective than a large number of closely located relay nodes with correlated channel coefficients. Also, the use of multiple relay nodes increases the cost and the transmitted power. In [24], a repetition/retransmission

Wireless Systems with New Cooperative Relaying Algorithm

281

procedure based on a cooperative protocol is proposed under the assumption that each retransmission is carried out on orthogonal channels. In this case, all available relay nodes are involved. However, since the available power is uniformly/linearly, but not optimally, distributed across all relay nodes, a scheme for the full participation of all nodes cannot be optimal. In a wireless environment, power is a limited resource, therefore, will be considered as the main limitation. Therefore, the selection of a single relay in a network with multiple potential relay nodes is a promising method that improves system performance [25]. The central controller is responsible for choosing the optimal relays for cooperation with centralized mechanism [21, 22], so it needs all the necessary information to select the best relay. Therefore, the nodes need to redirect their local information, for example, channel state, residual capacity, to the central controller. Such a centralized mechanism often requires a large overhead of signaling due to feedback messages from cooperating relay nodes to the central controller. Also, the state of the channels may change during the time of making decision and answers, that makes the choice non-optimal. Full CSI information is required or, at least, channel statistics are necessary for centralized schemes of relay selection. The advantages of this approach include the availability of complete network information from the central controller for optimal selection, and as a result, to achieve a higher level of performance than distributed analogues [26]. In [27], a relay selection scheme is described, based on channel state information (CSI) at the source and relay node. This scheme has a high complexity, since it requires the CSI of all participating communication lines and, therefore, finding the “best” relay nodes is necessary at each transmission time. The repeater is selected by the source or target receiver among all potential nodes that correctly received the packet from the source and can relay the message with a distributed mechanism [28]. Then the target receiver evaluates the message packets and selects the node that provides the largest cooperation gain as a relay. In [29], a distributed relay selection scheme was proposed to reduce the drawbacks of centralized approach, that use instantaneous SNR values. In [30], a scheme was proposed for determining the best relay node, which depends either on the minimum or on the harmonic average value of the SNR of the communication links source-relay and relay-target receiver. However, these schemes lead to a non-zero probability of choosing two relays for the same source, and the analysis presented in [29] gives a quantitative estimate of this probability. In [31], the efficiency of cooperative diversity is estimated when the best relay is selected based on the average SNR value and the probability of not choosing a relay node based on the instantaneous SNR value. As mentioned above, the task of optimal relay selection is to find the best one in dynamic environments of wireless networks, where the network topology may be changed, and the wireless environment changes over time due to the dynamic nature of such networks. So, to solve this problem it is necessary to determine the optimality criterion, i.e. parameter or parameters fir determination the “best relay” and actually find the relay node for a given set of channel conditions. The selection criterion for minimizing the probability of deny can be the value of received instantaneous SNR of the source-relay and relay- target receiver channels. However, obtaining instant SNR values for all potential Rs to find the current best can

282

L. Afanasieva and S. Kravchuk

result to significant overhead. With a given target data transfer rate between the source and the target receiver, there is no need to find the best R—it is enough to choose one that is likely to support a successful communication attempt [32]. The selection criterion is the value of average received SNR, when R is chosen for a longer period, for example, for the total network lifetime. The average SNR is equivalent to the scheme for choosing the distance between the relay and the target receiver [33]. In [34], the distance from the relay to the destination is taken as the main criterion for selecting a relay node. A cooperative protocol based on distance measurement chooses the node closest to the destination as the best relay. The disadvantage of this protocol is that it does not take into account the SNR value of received signals, which at the same distances can differ greatly due to interference, shadowing and the effects of multipath fading in wireless communication links. In [35], the best repeater is determined due to the information of the residual energy of the nodes in order to maximize the total lifetime of the network. That is, nodes with a low residual energy level are less likely to be chosen as a relay node for transmission. Relay selection schemes for cooperative protocols play an important role in cooperative communication systems, have a significant impact on diversity gain, and network performance. The selected optimal relay should be the best among all the candidates for relaying that can make the maximum contribution to improving network performance in terms of the probability of packet failure and channel utilization efficiency [36, 37]. In most of the works above, a fixed power level at the source and relay is assumed. Therefore, the aim of this work is to develop a new relay selection method for cooperative relaying, which takes into account the position of the node, radio channel parameters (signal-to-noise ratio and bit error rate) and energy criterion.

4 System Model The system model of a half-duplex wireless cooperative relay network, which consists of the source S, the target receiver D, and a set of N potential relay nodes Ri, is shown in Fig. 5. In this communication model, the “best” relay is selected from a set of potential relay nodes, which is then used to facilitate communication between the source and target receiver. It is assumed that all nodes are equipped with one antenna; the spectral power density of the noise is equal to N 0. All nodes can adjust their instantaneous transmit power in the range [0, Pmax ], where a certain physical limitation is imposed on Pmax . All relays amplify and transmit the information signal without any further processing, that is, they operate in the non-regenerating Amplify-and-Forward mode. It is supposed to use a slowly varying Rayleigh quasistatic channel with flat fading for each link, which consists of one direct (S → D) and two- hop channel (S → R, R → D). It was assumed that the transmission coefficients of the channels H (channel matrix) is fixed within the coherence time interval. Coherence time interval is meant the length of time during which the channel impulse response is assumed constant. Namely, the signal x(t) was sent at time t1 on the receiving side will be y1 (t) = x(t − t1 ) ∗ ht1 (t), where ht1 - channel impulse response at time t1 .

(1)

Wireless Systems with New Cooperative Relaying Algorithm

283

Fig. 5. The system model

Similarly, a signal transmitted at time t2 will be received as y2 (t) = x(t − t2 ) ∗ ht2 (t)

(2)

If the difference ht1 (t) − ht2 (t) is relatively small, then the channel can be considered constant within the interval [t1 , t2 ]. At the beginning of each coherent time interval, some pilot signals are transmitted to obtain all the necessary information about the channel state. Thereafter destination selects a relay with best channel condition according to adjusted algorithm and transmits the identifier value of the selected R to all potential devices. The selected relay will be used for transmission of information during the coherence time interval. Due to half duplex restrictions (R cannot “listen” and transmit simultaneously) each transmission period is divided into two time intervals: source – target receiver (S → D), source – relay (S → R) and relay – target receiver (R → D), respectively. Consequently, at the end of the second time interval, the terminal of the target receiver combines the signal received directly from the source with the signal that is received through the relay using the maximum ratio combining (MRC). It is assumed that the channel state information is known at the target receiver, and the relays do not participate in the transmission if the direct transmission from the source to the destination is successful. In case of unsuccessful direct transmission, the “best” relay transmits data received from the source to the target receiver during the second phase of the transmission, and the destination node receives the data. During the first phase, the source transmits its data xs to i -th relay R, where i = {1, 2, 3 . . . N } , and to the target receiver D. The received signal on D and on the selected R, is the attenuated signal from the source plus noise ySD = hSD xs + ns

(3)

ySRi = hSRi xs + nSRi , i = 1, 2, . . . N

(4)

where hSD , hSRi - are mutually independent values of the source – target receiver (S → D), source – relay (S → Ri )) and relay – target receiver (Ri → D), respectively.; nSRi , ns - noise in the relay and target receiver.

284

L. Afanasieva and S. Kravchuk

Then, the repeater amplifies the received data with the amplification factor β and redirects to the target receiver on the channel with hRi D . The received signal yD at the target receiver D can be written as yD = hSD xs + βhSRi hRi D xs + βhRi D nr + ns

(5)

In the case of instantaneous power limiting, it is required that β2 ≤

PRi

 2 Pnoise + Ps hSRi 

(6)

Assuming that the maximum ratio combining method is used in the node of the target receiver, the total SNR in the target receiver can be written as SNRcoop = SNRsd + (SNRsr SNRrd )/(SNRsr + SNRrd + 1)

(7)

where SNRsr = PS |hSR |2 /N0 and SNRRD = PR |hRD |2 /N0 is the signal-to-noise ratio at the input of the repeater receiver and the target receiver, respectively; N 0 is the intrinsic noise power in the passband of the receiver.

5 Proposed Selection Algorithm 1. Based on the requirements of predefined minimum value SNRmin that is required for correct decoding of the message on R (i.e., with a maximum allowable number of bit errors), determine from the set of terminals A, that are in the coverage area S, the set N of potential relay nodes (A ⊂ N), that satisfy this requirement   (8) N ∗ = Ri ∈ N|SNRSRi ≥ SNRmin 2. Determine the maximum value of the radius of the location of the set N of potential relay nodes. Source S is located at a distance d from the target receiver (Fig. 1). The value of rmax , is determined from the expression rmax = f

10γSR /10∗α 4π

(9)

where α - pathloss coefficient for urban development is 3, a γSR - attenuations from S to R and are found from the expression γSR = 10ln

PSX ∗ GTX ∗ GRX SNRmin ∗ Pnoise

(10)

3. From the set of N relays, choose the one that has the lowest attenuation value of the Ri → D. The source uses the metric μr corresponding to the expression   μr = min γRi D (11) i∈N

Wireless Systems with New Cooperative Relaying Algorithm

285

The mode of operation is described in the following step-by-step algorithm:

Consider the analysis of the effectiveness of the cooperative diversity method with the Amplify and Forward relaying protocol when choosing the best relay according to the proposed algorithm. Performance characteristics, such as the number of bit errors, were evaluated and compared with the direct channel transmission method with Rayleigh attenuation. The parameters required for analysis, SNRmin = 20 dB for the link → Ri , the distance between the source and target receiver d = 150 m. In Fig. 6 compares the BER characteristics of the binary phase modulation (BPSK) in proposed algorithm with the direct transmission scheme. It should be noted that the proposed scheme, based on threshold indicators, surpasses the direct transfer scheme in BER performance.

Fig. 6. The BER performances for direct and cooperative relay transmissions

286

L. Afanasieva and S. Kravchuk

Since the proposed algorithm requires several steps to select the best relay, it has greater efficiency BER, as well as the gain in power consumption for signal transmission.

6 Conclusions The process of choosing a method of cooperative transmission is important to maintain a high level of quality of service, however, the selection of a retranslation node for implementing cooperative data transfer is also an important task. A new algorithm has been developed to select the best relay, it considers a number of criteria, namely: the position of the relay, the bit error rate, power consumption in the relay nodes. A system model of a wireless system using this algorithm was built in the MATLAB software environment. During the simulation the evaluation and comparison of the characteristics of schemes with direct transmission, opportunistic transmission and transmission using cooperative relaying in terms of bit error rate and throughput. The simulation results showed that the proposed best relay algorithm shows significant advantages compared with the given transmission scheme. The analysis of the data obtained as a result of modeling showed the following. The cooperative relay mechanism allows to increase the value of SNR, and the more degradation (attenuation) on the radio link between the eNodeB and the MS, the better gain from the inclusion of the cooperation scheme. It was shown that the decision to use cooperative relaying should occur when a certain distance between the eNodeB and MS is reached (approximately d = 0.7–0.8), when the cooperation gain is about 5–7 dB.

References 1. Ilchenko, M.Y., Kravchuk, S.O.: Telecommunication Systems. Naukova dumka, Kiev (2017) 2. Dahlman, E., Parkvall, S., Skold, J.: 5G NR: The Next Generation Wireless Access Technology. Academic Press, London (2018) 3. Kramer, G., Mari´c, I., Yates, R.D.: Cooperative Communications. Found. Trends Netw. 1(3– 4), 271–425 (2007) 4. Romanov, O.I., Hordashnyk, Y.S., Dong T.T.: Method for calculating the energy loss of a light signal in a telecommunication Li-Fi system. In: Proceedings of the 2017 International Conference on Information and Telecommunication Technologies and Radio Electronics, UkrMiCo, pp. 1–7. IEEE Conference Publications, Odessa, Ukraine (2017) 5. Ray Liu, K.J., Sadek, A.K., Su, W., Kwasinski, A.: Cooperative Communications and Networking. Cambridge University Press, Cambridge (2008) 6. Zakharov, A., Rozenko, S., Ilchenko, M.: Varactor-tuned microstrip bandpass filter with loop hairpin and combline resonators. IEEE Trans. Circ. Syst. II Exp. Briefs 66(6), 953–957 (2019) 7. Song, W., Ju, P., Jin, A.-L.: Protocol Design and Analysis for Cooperative Wireless Networks. WN. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-47726-8 8. Tang, Z.Z., Wang, H.Y., Hu, Q.: An energy-efficient relay selection strategy based on optimal relay location for af cooperative transmission. In: IEEE International Symposium World of Wireless, Mobile and Multimedia Networks (WoWMoM), San Francisco, CA, pp. 1–4 (2012) 9. Zakharov, A., Rozenko, S., Litvintsev, S., Ilchenko, M.: Trisection bandpass filter with mixed cross-coupling and different paths for signal propagation. IEEE Microw. Wirel. Compon. Lett. 30, 12–15 (2020)

Wireless Systems with New Cooperative Relaying Algorithm

287

10. Asshad, M., Khan, S.A., Kavak, A., Küçük, K., Msongaleli, D.L.: Cooperative communications using relay nodes for next-generation wireless networks with optimal selection techniques: a review. IEEJ Trans. Elect. Electron. Eng. 14(5), 655–805 (2019). Institute of Electrical Engineers of Japan, John Wiley & Sons, Inc. 11. Romanov, O., Mankivskyi, V.: Optimal traffic distribution based on the sectoral model of loading network elements. In: Proceedings of the 2019 IEEE International Scientific-Practical Conference Problems of Infocommunications, Science and Technology (PIC S&T), Kyiv, Ukraine, pp. 683–688 (2019) 12. Afanasieva, L.A., Kravchuk, S.O., Minochkin, D.A.: Outage probability in wireless communication system with multiple antenna cooperative retransmission. Sci. Pap. Coll. Mil. Inst. Kyiv Natl. Univ. Shevchenko 51, 19–26 (2016) 13. GPP TS 23.203 V12.6.0: Technical Specification Group Services and System Aspects; Policy and charging control architecture (2014) 14. Yuan, Y.: LTE-Advanced Relay Technology and Standardization. Springer, Berlin (2013). https://doi.org/10.1007/978-3-642-29676-5 15. Afanasieva, L.O., Kravchuk, S.O.: Adaptive relay protocol for cooperative networks. In: Dig. of the 13th International Scientific conference on Modern Challenges in Telecommunications, Kyiv, Ukraine, pp. 192–194 (2019) 16. Romanov, O.I, Oryschuk, M.V., Hordashnyk, Y.S.: Computing of influence of stimulated Raman scattering in DWDM telecommunication systems. In: Proceedings of the 2016 International Conference Radio Electronics and Info Communications, UkrMiCo, pp. 1–4. IEEE Conference Publications, Odessa, Ukraine (2016) 17. Afanasieva, L.O., Kravchuk, S.O, Shevchenko, I.I., Rystsova, K.I.: Coded cooperation scheme. In: Dig. of the 12th International Scientific Conference on Modern Challenges in Telecommunications, Kyiv, Ukraine, pp. 211–213 (2018) 18. Rizhko, A.V., Kravchuk, S.O.: Investigation of ability to increasing quality of service for IEEE 802.16j multihop repeater networks. In: Dig. of the Eleventh International Scientific Conference on Modern Challenges in Telecommunications, 18–21 April 2017, Kyiv, Ukraine, pp. 163–165 (2017) 19. Lei, X., Hong-Wei, Z., Xiao-Hui, L., Xian-Liang, W.: Optimum relay location in cooperative communication networks with single AF relay. Int. J. Commun. Netw. Syst. Sci. 4, 147–152 (2011) 20. Ismail, M., Zhuang, W.: A distributed multi-service resource allocation algorithm in heterogeneous wireless access medium. IEEE J. Sel. Areas Commun. 30, 425–432 (2012) 21. Li, Y., Wang, P., Niyato, D., Zhuang, W.: A dynamic relay selection scheme for mobile users in wireless relay networks. In: 30th IEEE International Conference on Computer Communications, Joint Conference of the IEEE Computer and Communications Societies, Shanghai, China (2011) 22. Du, Q., Zhang, X.: QoS-aware base-station selections for distributed MIMO links in broadband wireless networks. IEEE J. Sel. Areas Commun. 29(6), 1123–1138 (2011). IEEE 23. Shan, H., Zhuang, W., Wang, Z.: Distributed cooperative MAC for multihop wireless networks. IEEE Commun. Mag. 47(2), 126–133 (2009) 24. Laneman, J.N., Tse, D.N., Wornell, G.W.: Cooperative diversity in wireless networks: efficient protocols and outage behavior. IEEE Trans. Inform. Theor. 50, 3062–3080 (2004) 25. Jing, Y., Jafarkhani, H.: Single and multiple relay selection schemes and their achievable diversity orders. IEEE Trans. Wirel. Commun. 8(3), 1414–1423 (2009) 26. Yang, Y., Hu, H., Xu, J., Mao, G.: Relay technologies for WiMax and LTE-advanced mobile systems. IEEE Commun. Mag. 47(10), 100–105 (2009) 27. Ibrahim, A.S., Sadek, A.K., Su, W., Liu, K.J.R.: Cooperative communications with relayselection: when to cooperate and whom to cooperate with? IEEE Trans. Wirel. Commun. 7(7), 2814–2827 (2008)

288

L. Afanasieva and S. Kravchuk

28. Beres, E., Adve, R.: On selection cooperation in distributed networks. In: 40th Annual Conference on Information Sciences and Systems, Princeton, US-NJ, pp. 1056–1061 (2006) 29. Bletsas, A., Lippman, A., Reed, D.: A simple distributed method for relay selection in cooperative diversity wireless networks, based on reciprocity and channel measurements. In: Proceedings of IEEE 61st Vehicular Technology Conference, Stockholm, Sweden (2005) 30. Michalopoulos, D.S., Karagiannidis, G.K.: Performance analysis of single relay selection in Rayleigh fading. IEEE Trans. Wirel. Commun. 7, 3718–3724 (2008) 31. Bletsas, A., Shin, H., Win, M., Lippman, A.: Cooperative diversity with opportunistic relaying. In: IEEE Wireless Communications and Networking Conference, WCNC 2006, Las Vegas, NV, USA, vol. 2, pp. 1034–1039 (2016) 32. Bletsas, A., Khisti, A., Reed, D.P., Lippman, A.: A simple cooperative diversity method based on network path selection. IEEE J. Sel. Areas Commun. 24(3), 659–672 (2006) 33. Eshteiwi, K.M.: Outage performance of relay selection in cooperative wireless networks over Rayleigh fading channels. In: 26th Annual IEEE Canadian Conference on Electrical and Computer Engineering, CCECE 2013, pp. 1–5 (2013) 34. Chen, M., Liang, X., Leung, V., Balasingham, I.: Multi-hop mesh cooperative structure based data dissemination for wireless sensor networks. In: The 11th International Conference on Advanced Communication Technology, ICACT’09, Phoenix Park, Korea, pp. 102–106 (2009) 35. Chen, Y., Yu, G., Qiu, P., Zhang, Z.: Power-aware cooperative relay selection strategies in wireless ad hoc networks. In: IEEE International Symposium on Personal, Indoor and Mobile Radio Communication (PIMRC), Helsinki, Finland (2006) 36. Su, Y., Li, Y., Wu, X., Liu, L.: Outage performance for amplify-and-forward two-hop multipleaccess channel with noisy relay and interference-limited destination. IET Commun. 12, 205– 213 (2018) 37. Afanasieva, L., Minochkin, D., Kravchuk, S.: Providing telecommunication services to antarctic stations. In: Proceedings of the 2017 International Conference on Information and Telecommunication Technologies and Radio Electronics, UkrMiCo 2017. IEEE Conference Publications, Odessa, Ukraine (2017)

Autonomous Unmanned Aerial Vehicles Communications on the Base of Software-Defined Radio Mykola Kaidenko(B)

and Sergey Kravchuk(B)

Department of Telecommunications, National Technical University of Ukraine “I. Sikorsky KPI”, Kyiv, Ukraine {kkk610,sakravchuk}@ukr.net

Abstract. The principles of the functional and structural construction of a communication part of the UAV system based on SDR and SoC technologies while ensuring the survivability of the formed radio channels are presented. The main increase in survivability is achieved through the use of adaptation over the frequency range using an additional reception channel for continuous analysis of interference conditions, using two or more data transmission channels, complex algorithms for optimal selection of the operating range, transmission channel parameters and adaptive protocols for simultaneous data transmission over two or more channels. Keywords: Unmanned aerial vehicles · Communication system · Communication channels · SDR and SOC technologies

1 Introduction Currently, a significant rise is undergoing the creation of unmanned aerial vehicles (UAVs) for a variety of applications. There was a wide range of UAVs - from heavy strategic to light small (mini and micro), from remotely piloted vehicle to fully autonomous. The number of UAVs is steadily increasing and the range of their tasks is expanding, especially related to the protection, control and monitoring of facilities, the elimination of emergency situations, the acquisition of spatial data and military applications [1–3]. At the same time, the development of interoperability, reliability, efficiency and survivability of the components of the UAV is a prerequisite for the efficient operation of electronic systems placed on board in a limited space [4]. Prospective onboard UAV equipment complexes, as well as onboard equipment complexes of prospective manned aircraft, should consist of the following interconnected systems: management information system, navigation equipment, onboard radio system, target load systems, radar identification system. The UAV communication system has its own distinctive channel characteristics in comparison with the widely used cellular or satellite systems. The parameters of the communication channel and its survivability (the ability to transmit information with a © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 289–302, 2021. https://doi.org/10.1007/978-3-030-58359-0_16

290

M. Kaidenko and S. Kravchuk

given probability of error) of the channel is crucial for the effective management of the UAV and the performance of its tasks. Nevertheless, there are several key problems in creating communication channels with UAVs. These problems can be divided into two classes. The first class of problems is related to the fact that the communication channel with the UAV is non-stationary, in which spatial and temporal changes occur and unintended interference in the form of fading caused by multipath and shading occurs. The second class of problems is related to the fact that, in parallel with the development of the UAV, methods and means of purposeful disruption of their normal functioning are being developed, through the creation of deliberate interference. Modern technical means allow the detection and direction finding of control channels and the discharge of information from a UAV, to interfere with the operation of on-board equipment and ground-based control complexes. The main risk factors due to intentional interference should include a violation of the control of the UAV due to: impact on the communication channel in order to destroy it; impact on the communication channel in order to intercept control of the UAV, including unauthorized access to the main nodes at the program level, leading to disruption of the control of the operator. Thus, in the current requirements for the development of modern UAV systems, the further increase in the efficiency of the communication equipment, which forms the radio link for data transmission both for control and for maintenance of the airborne load, remains relevant. This is also confirmed by numerous publications on this topic. So, in [5], a generalized analysis of the importance of the fact that UAVs must be stable when transmitting data and reliable in flight is carried out. At the same time, the problem of information security of data transmission remains relevant. The article [6] describes the possibilities of remote monitoring of the flight parameters of the UAV and its payload data in real time. Methods and technologies for transferring remote control and monitoring parameters are proposed, a general hardware model for data transmission and a software model of a communication system suitable for a UAV are presented. Improve accuracy and reliability of data transmission in the communication channel between the transport UAV ground control station and discussed in [7]. In [8], an analysis of the characteristics of UAV data wireless channels was carried out using the flat fading and selective fading models. However, these models are suitable for low-speed data channels. In [9] presents aspects of construction and methods of implementation of the aero-stratospheric telecommunications platform in Ukraine for example HAPS (High Altitude Platform Station) studies conducted by the Igor Sikorsky Kyiv Polytechnic Institute. In [10] the general structural and functional principles of building a control and communication system for on-board and ground equipment of a UAV-based telecommunication network are developed. At present, advanced technologies, in particular, SDR (Software-Defined Radio) and SoC (System-on-chip) enable developers to improve spectrum utilization, open new telecommunications markets and expand the list of available functions, in particular the Cognitive Radio System. The use of approaches used in the Cognitive Radio System to create a communication system for UAVs will allow to solve the problems of adapting a radio system operating under intentional interference [11].

Autonomous Unmanned Aerial Vehicles Communications

291

Thus, when building a UAV communication system, it is necessary to take into account that it should work effectively in conditions of unintentional and intentional interference and be adaptive to changing working conditions. The implementation of such requirements is possible only when using the technologies of SDR and SoC. Therefore, the purpose of this work is to develop a functional-structural construction of a communication part of the UAV system based on SDR and SoC technologies with the provision of increasing the survivability of the radio channels formed.

2 Structural and Functional Principles of Building Onboard Equipment of a Communication and Control System UAV The control and communication system of the UAV is considered as a single set of equipment connected by common tasks. Structurally functional principles of building the system take into account the fact that it should provide continuous data exchange via the control channel and telemetry with high reliability, as well as continuous data transmission over the data channel towards the ground-based system with guaranteed quality and, accordingly, should be adaptive for work under conditions of intentional and unintentional interference. The overall structure of the onboard control and communication system is shown in Fig. 1.

Fig. 1. The structure of the UAV control and communication system

The control channel and telemetry are used to transfer UAV control data from the ground station and telemetry data from the aircraft to the ground station. To ensure continuity of management of the aircraft at all stages of flight using an antenna with circular radiation pattern. For the same purpose, the frequency range is 70 to 900 MHz. The data

292

M. Kaidenko and S. Kravchuk

transfer rate in the channel is relatively low up to 100 kbps. Ensuring noise immunity and protection from deliberate interferences is carried out using spectrum expansion technology: for direct use - DSSS (Direct Sequence Spread Spectrum); for military use FHSS (Frequency-Hopping Spread Spectrum). Modulation FM4, interference-encoding - cascading code, or block turbo coding with variable coding speed. The allowable data transmission delay in the channel is determined by the appointment of the UAV and the speed of its movement, respectively, and the necessary response time to the control command and is from 10 to 200 ms. The use of SDR for constructing a control channel can additionally provide the possibility of choosing the optimal carrier frequency in the range of operating frequencies, which can provide additional protection against target intentional noise by using frequency adaptation algorithms. For data transmission channels are used in higher frequency ranges from MHz to 6 GHz, and above. In order to ensure the guaranteed quality of data transmission in the context of deliberate interference, it is expedient to use two SDR transceivers. One is constantly working, and the other one is used to analyze the entire frequency range in order to find the optimal transmission speed and the absence of broadband intentional interference, including in the frequency band of the control channel and telemetry. In case of deliberate disturbance detection SDR transceivers are changing places. Due to the fact that data channels are high-speed and use a deliberate control mechanism, these channels can be used additionally as duplicates for the control channel and telemetry, or for transmitting control and telemetry data simultaneously on two different routes. Such an approach further provides the channel’s stability from intercepting and simulating control signals and telemetry, since in order to intercept the UAV control, it will be necessary not only to overcome the cryptographic protection at the same time in two channels, but also to use synchronization simultaneously on two channels, with one of which is “wandering”. Data channels are asymmetric with time duplex, the channel towards the ground station is high-speed, the line down/line up can be 8/1, 16/1. Modulation in transmission channels - multi-position PSK2, QPSK, QAM16, QAM64, QAM256. Interferenceencoding - cascading code, block turbo code or LDPC code with variable coding speed. Adaptive modulation and encoding are used to provide for the control of unintended interference. For the frequency adaptation, adaptive modulation and coding, the principles and algorithms described in [12, 13]. For data transmission, an antenna system with directional antennas is used: a set of planar antennas, or a mirrored antenna with several range desalinizes. Antenna system includes additional elements for controlling the rotation of antennas in two azimuthal planes to provide work in the mode of automatic tracking of the direction to the ground station. To ensure the antenna position remains unchanged, it is positioned on a gyrostabilized platform. The functions of hydrostabilization are provided with an additional turning mechanism (third plane for compensating the platform roll) and a system of three coordinate sensors MEMS. The UAV’s useful load is a set of cameras for video surveillance: a PTZ camera (Pantilt-zoom) - a camera that supports remote control of the direction and zoom, as well as (if necessary) an infrared camera or night vision camera. These two (three) cameras

Autonomous Unmanned Aerial Vehicles Communications

293

are installed on an independent gyrostabilized platform. Another camera - the course camera is used for visual observation of the flight process by the UAV operator and manual flight control, the camera is a wide-angle camera with relatively low resolution. The camera is stationary in the direction of UAV flight. Autopilot, UAV flight control devices and telemetry sensors provide flight control and determine the necessary parameters for this. In order to determine the orientation of UAV in space, the GPS system (Galileo, Glonass) is used to determine the coordinates of the UAV and, as a duplicate, the SINS (strapdown inertial navigation systems). Despite the fact that the accuracy of the SINS based on MEMS sensors is relatively low, it can be used to provide an orientation of the UAV in the absence of (deliberate suppression) of GPS signals. The On-Board Data and Data Processing and Switching and Routing Unit is organized on the basis of SoC-system technology, which combines the HPS (Hard Processor System) and FPGA (Field- Programmable Gate Array). The on-board processor for data management and data processing is based on the Linux operating system using HPS SoC, and provides the following functions: support for the on-board system required for the original drivers for interfacing with external SoC interfaces; calculation of data from MEMS gyrostabilized platform sensors, calculation of necessary corrections and management of the platforms; processing and analysis of data for orientation of UAVs in the space coming from GPS and SINS systems, as well as data from telemetry sensors, decision-making on choosing the optimal system for determining the orientation of the UAV; formation and processing of management data and telemetry data packets; analysis of data on the state of the control channel and telemetry, data transmission channels, decision making on optimization of traffic routes; support of the MPTCP (Multipath TCP) technology to ensure a stable connection between the UAV and the ground station in the event of intentional and unintentional disturbances; processing and compressing video and images from cameras for transmission over communication channels, with the compression ratio adaptive and depends on the available channel resource at a specific time; managing the work of the routing and switching unit; exchange of data with an autopilot and record logs (documentation) of the flight; logging of the operation of the control and communication system; recording information from video surveillance cameras in the event that data cannot be transmitted. The switching and routing unit is implemented on the FPGA SoC and operates under the control of the onboard processor for data management and data processing. FPGA features functions such as packet generation, high-speed routing and switching, support for hardware interfaces and data bus with different speeds and different bit rates. In addition, some functions that require high-speed computing, as well as functions for forming control signals for gyrostabilized platforms can be carried over.

3 Structural and Functional Principles of Building Ground Equipment of a Communication and Control System UAV Structural and functional principles of the construction of a ground control system and communication take into account the designation of the UAV, the structure of the construction of the onboard control system and communication, and the need to ensure

294

M. Kaidenko and S. Kravchuk

the survivability of the UAV in the event of different types of interference. The overall structure of the UAV ground control and communication system is shown in Fig. 2. The system consists of: control channel and telemetry with an antenna of the corresponding type; data transmission channels; antenna system of automatic maintenance of UAV; on-board processor for data processing and management; block of switching and routing; display system from video surveillance cameras; flight control and reflection from the course camera, as well as the system for determining the coordinates of the ground station. The communication channels of the terrestrial system have the parameters corresponding to the communication channels of the UAV. As an antenna of the control channel and telemetry, a wave channel (logo periodic antenna) or an antenna with a circular orientation diagram can be used; it is attached to the antennas of the data transmission system with the automatic maintenance of the UAV. As antennas, mirrored antennas with multiple range lights are used. The diameter of the mirrors is determined by the tactical purpose of the UAV and the maximum range of flight. The antenna system should provide automatic maintenance of the UAV during the entire flight. The functions and structure of the processor for data management and data processing, switching and routing block are similar to those that are installed on the UAV. The processor of data management and data processing additionally performs calculations for the antenna system of automatic maintenance, calculation of the flight path of the UAV, the definition of necessary amendments, logging (documentation) of the flight process and the exchange of data from the UAV.

Fig. 2. The structure of the UAV control and communication system

Autonomous Unmanned Aerial Vehicles Communications

295

To control flight UAV uses the flight control panel together with the manual control device. Images from the course camera are displayed on the flight control monitor along with other information that is required for flight monitoring and management. To display images from video surveillance cameras, individual monitors are used, which can display not only video information but also topographic maps. For processing video and cartographic information, an image processing engine is used, which is executed as a separate module.

4 Technical Requirements and Control Algorithms for On-Board and Terrestrial Antenna Systems The technical requirements for a terrestrial antenna system are determined based on the following: the antenna system should perform the functions of automatic tracking of the aircraft; automatic tracking of the aircraft is used for data transmission channels to a ground station and control and telemetry channels; the range of automatic tracking of an aircraft is determined by its tactical tasks; automatic tracking is carried out throughout the flight of the aircraft with a preliminary installation of the source data; the antenna system consists of several antennas for various frequency ranges and various purposes (control and telemetry channel, data transmission channels); the antenna system should be designed to support MIMO technology for data channels; the antenna system should provide maximum energy efficiency of the communication channel. Technical requirements for a terrestrial antenna system: – range of automatic tracking of the aircraft: in azimuth 360°; in elevation 0–+60°, which corresponds to a range of 17 km with a flight altitude of 30 km; – with automatic tracking in azimuth, the 360° range provides for one revolution of the antenna system with fixation of the extreme position; – the type of antenna of the data transmission channel is mirrored with several irradiators for different frequency ranges, or planar for each range; – type of antenna of the control channel and telemetry - wave channel, or log-periodic; – placement of the antenna system - a tripod, or on a movable chassis; – placement of transceiver equipment - on the antenna system; – operating mode - automatic maintenance, manual installation; – to ensure the automatic tracking mode, a separate module is used, connected by interfaces to the rotary mechanisms of the antenna system, a ground station automatic positioning system, a control and telemetry channel, and a data transmission channel; – rom the control and telemetry channel, the automatic tracking module receives data from the aircraft positioning system; – from the data transmission channel, the automatic tracking module receives data on the level of the received RSSI signal (Received Signal Strength Indication) and the effective signal-to-noise ratio in the SNR (Signal-to-Noise Ratio) channel. The control algorithms for the ground-based antenna system are developed using data from the aircraft positioning system and the ground-based system, as well as based on energy estimates of the received signal.

296

M. Kaidenko and S. Kravchuk

The initial initialization algorithm is shown in Fig. 3. During initial initialization, data from the aircraft positioning system and the ground system are used. Before initialization, the antenna system is installed in the direction of UAV take-off to block the extreme position of the antenna system.

Fig. 3. Initial initialization algorithm of antenna-system systems with automatic tracking of an aircraft

The compass of the antenna system is calibrated as follows: – – – –

the antenna system is installed in position +30° in elevation; the antenna system rotates in azimuth of −90°s, +180°, −90°; a change in the position of the antenna system in elevation +30−60°; indicators of the MEMS (microelectromechanical systems) compass, accelerometer and gyroscope are recorded in the memory of the automatic tracking module;

Autonomous Unmanned Aerial Vehicles Communications

297

– calculations of correction indicators “hard metal” and “soft metal” are carried out, the formulas for correction differ for each type of sensors and are determined according to the documentation; – Correction indicators are entered into the memory of the automatic tracking module for further making the system. It is assumed that the accuracy of the compass calibration is estimated by the indicator and compared with the compass, which is at least 10 m away from the metal objects. In order to obtain data on UAV coordinates, the control and telemetry channel must be activated at the minimum transmitter power (determined by the required signal-to-noise ratio). In case the UAV coordinates are not received, the control channel or interface with the control channel is defective. After identifying the cause, the initialization is repeated. Verification of the initial initialization is performed visually according to the indicators that are displayed on the monitor: UAV range, azimuth, and position angle. If the initialization is not performed, the antenna system must be checked for the rotary mechanisms. The automatic tracking algorithm in flight mode is shown in Fig. 4. The mode is used as the main one when the aircraft positioning system data is available. After the flight of the aircraft, the antenna system is transferred to the main mode of operation. The coordinates of the antenna system and the current coordinates of the UAV are recorded in the memory of the automatic tracking system. In flight mode, UAV coordinates are constantly updated. To eliminate “false” data and predict the next location of the UAV, filtering is performed with the calculation of the trajectory of its movement. As a filter, a linear interpolation filter interpolator can be used. For a more accurate determination of the UAV motion path, it is advisable to use the Kalman filter, which will additionally provide predictions of the UAV motion in the absence of data on its coordinates. The distance to the UAV is calculated in order to determine the permissible deviations in azimuth and elevation for the automatic tracking system, to estimate the energy reserve in the data transmission channels and, based on it, to indirectly assess the presence of energy interference, to determine when the automatic tracking system shuts off during UAV landing. The need to change the position of the antenna system is determined as a function of the width of the narrow radiation pattern of all available antennas that will be involved in individual flight segments. At a distance to the UAV of more than 1 km, the permissible deviations from the exact azimuth and elevation angle are up to half the width of the radiation pattern in the corresponding plane. At a distance of less than 1 km, deviations can exceed this indicator, in addition, the energy reserve in the communication channels involved at this stage of the flight is taken into account. If necessary, the antenna system rotates in two angular coordinates. Data exchange on UAV coordinates is carried out constantly, in the absence of data from it and when determining how an additional indicator can be used to estimate the signal-to-noise ratio in a channel (estimates from all channels can be used). The automatic tracking algorithm based on energy estimates of the received signal (Fig. 5) is used in flight mode as auxiliary, or as an alternative in the absence of GPS data of the aircraft’s positioning system, while the UAV does not have strapdown inertial navigation system, or its data gives significant error. The transition to this algorithm is

298

M. Kaidenko and S. Kravchuk

Fig. 4. Algorithm for automatic UAV tracking in flight mode

automatically carried out in the absence of data on the coordinates of the UAV. The lack of coordinate data can be caused by two factors: the system for determining coordinates by GPS navigator indicators does not work; due to intentional interference, the GPS navigation system gives false UAV coordinates. In addition, during automatic tracking based on energy ratings, it is possible to assess the position of the UAV. The position of the UAV is estimated with the following accuracy: according to the angular coordinates, the accuracy is half the width of the antenna pattern in the corresponding plane, which was laid down in the algorithm; in terms of range, the accuracy is determined by SNR and RSSI estimates recalculated into the range, considering the channel as a channel with additive white Gaussian noise. The essence of the algorithm is to determine the maximum of the objective function for azimuth ϕ and elevation θ: (ϕ, θ) = max(SNR(ϕ, θ)) and (ϕ, θ) = max(RSSI(ϕ, θ)).

Autonomous Unmanned Aerial Vehicles Communications

299

Fig. 5. Algorithm for automatic UAV tracking in flight mode based on energy ratings

The algorithm provides for the search for the maximum function first in azimuth and then in elevation. The speed of the algorithm is determined by the capabilities (speed) of the antennarotary device. The elevation scanning algorithm is similar to the azimuth scanning algorithm. During the operation of the algorithm, it is supposed to constantly check the availability of data on the UAV coordinates, in case they appear, a transition to the automatic UAV tracking algorithm in flight mode is performed.

300

M. Kaidenko and S. Kravchuk

5 The Software Structure of the Onboard Processor for Control and Data Processing, the Ground Computing Complex of Control and Data Processing The software is necessary for the on-board control and data processing processor, the ground control and data processing computer complex to comply with the structural and functional principles for constructing the equipment of the on-board control and communication system on the UAV and the ground control and communication system. The operating system used is a Linux-based operating system. Conventionally, software is divided into system, application and tool. The system software includes the operating system itself with a graphical shell, standard drivers and drivers for working with memory (supporting the specifics of work for SoC) and the necessary utilities of the operating system.

Fig. 6. The software structure of the on-board processor control and data processing

The software includes the necessary set of drivers for working with devices and systems external to SoC, as well as internal SoC exchange drivers. These drivers are not part of the operating system and must be developed separately. The approach using drivers is the most appropriate, since it is complex and allows you to use all the features of the operating system and FPGA when SoC works with peripheral devices. This approach also allows the use of a single driver for devices of the same type, which greatly simplifies the system development process. The special application software developed by us is used to perform the functions of management, data processing, data exchange. C, C++, Python languages can be used

Autonomous Unmanned Aerial Vehicles Communications

301

to create application software. The general structure of the software for the on-board control and data processor is shown in Fig. 6. The general structure of the software for the ground-based computer complex for data management and processing is shown in Fig. 7.

Fig. 7. The structure of the software of the ground computing complex for data management and processing

6 Conclusion The principles of the functional and structural construction of a communication part of the UAV system based on SDR and SoC technologies while ensuring the survivability of the formed radio channels are presented. The main increase in survivability is achieved through the use of adaptation over the frequency range using an additional reception channel for continuous analysis of interference conditions, using two or more data transmission channels, complex algorithms for optimal selection of the operating range, transmission channel parameters and adaptive protocols for simultaneous data transmission over two or more channels. At the same time, in quantitative terms, the survivability grows at least two to three times, since for the suppression of the control channel it is necessary to simultaneously use two or three electronic warfare stations.

References 1. Namuduri, K., Chaumette, S., Kim, J.H., Sterbenz, J.P.G. (eds.): UAV Networks and Communications. University Press, Cambridge (2017). ISBN: 9781107115309. 256 p.

302

M. Kaidenko and S. Kravchuk

2. Ilchenko, M., Kravchuk, S., Kaydenko, M.: Combined over-the-horizon communication systems. In: Ilchenko, M., Uryvsky, L., Globa, L. (eds.): Advances in Information and Communication Technologies: Processing and Control in Information and Communication Systems. Lecture Notes in Electrical Engineering, vol. 560. Springer, Cham, pp. 121–145 (2019). https://doi.org/10.1007/978-3-030-16770-7, https://doi.org/10.1007/978-3-030-16770-7_6 3. Ilchenko, M.Y., Kravchuk, S.O.: Telecommunication systems, p. 736. Naukova dumka, Kyiv (2017). ISBN: 978-966-00-1566-1. (in Ukraine) 4. Valavanis, K.P., Vachtsevanos, G.J., (eds.): Handbook of Unmanned Aerial Vehicles, vol. LXXIX, p. 3022. Springer, Dordrecht (2015). ISBN 978-90-481-9706-4, ISBN 978-90-4819707-1 5. Çuhadar, ˙I., Dursun, M.: Unmanned air vehicle system’s data links. J. Autom. Control Eng. 4(3), 189–193 (2016). https://doi.org/10.18178/joace.4.3.189-193 6. Hristov, G.V., Zahariev, P.Z., Beloev, I.H.: A Review of the characteristics of modern unmanned aerial vehicles. Acta Technologica Agriculturae 2, 33–38 (2016). https://doi.org/ 10.1515/ata-2016-0008 7. Atoev, S., Kwon, O.H., Lee, S.H., Kwon, K.R.: An efficient SC-FDM modulation technique for a UAV communication link. Electronics 7(352), 1–18 (2018). https://doi.org/10.3390/ele ctronics7120352 8. Yu, G.F., Wang, Z.H., Li, L., Yang, C., Liu, D.W., Wang, Y.T.: Study on wireless channel models of UAV data link. Appl. Mech. Mater. 543–547, 2605–2608 (2014). https://doi.org/ 10.4028/www.scientific.net/AMM.543-547.2605 9. Zgurovsky, M., Ilchenko, M., Kravchuk, S., Kotovskyi, V., Narytnik, T., Cybulskyi, L.: Prospects of using of aerial stratospheric telecommunication systems. In: Proceedings of the 2016 IEEE International Scientific Conference on RadioElectronics & InfoCommunications (UkrMiCo 2016), Kyiv, Ukraine, 11–16 September 2016, pp. 20–23. IEEE Conference Publications (2016). IEEE Xplore Digital Library. https://doi.org/10.1109/UkrMiCo.2016. 7739636 10. Ilchenko, M.Y., Kaydenko, M.M., Kravchuk, S.O.: Structural-functional principles of management and communication systems for border and landscape equipment of the telecommunication network on the basic of aeroplatform. In: Dig. of the 12th International Scientific Conference on Modern Challenges in Telecommunications, Kyiv, Ukraine, 16–20 April 2018, pp. 26–29 (2018) 11. Kaidenko, M.M., Roskoshnyi, D.V.: Software defined radio in communications. In: Ilchenko, M., Uryvsky, L., Globa, L. (eds.) Advances in Information and Communication Technologies. Lecture Notes in Electrical Engineering, vol. 560. Springer, Cham (2019). https://doi.org/10. 1007/978-3-030-16770-7_11 12. Kravchuk, S., Kaidenko, M.: Features of creation of modem equipment for the new generation compact troposcatter stations. In: Proceedings of the International Scientific Conference on RadioElectronics & InfoCommunications (UkrMiCo 2016), Kyiv, Ukraine, 11–16 September 2016, pp. 365–368. IEEE Digital Library (2016) 13. Kaydenko, N.: Adaptive modulation and coding in a broadband wireless access systems. In: 23th International Crimean Conference on Microwave and Telecommunication Technology, CriMico 2013, Sevastopol, 8–13 September 2013, pp. 275–276 (2013)

Estimation of the Motion Parameters of the UAVs FANET Using the Dynamic Filtering Method Oleg Tsukanov(B)

and Evheny Yakornov(B)

National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, 37, Prosp. Peremohy, Kyiv 03056, Ukraine [email protected], [email protected]

Abstract. The algorithm for estimating the motion parameters of unmanned aerial vehicles as elements of flying wireless sensor networks FANET is considered. To estimate the motion parameters, the proposed stable dynamic filtering algorithm allows us to obtain estimates of coordinates and their derivatives. The results obtained can simultaneously improve the accuracy of the estimation of motion parameters, ensure the sustainability of the assessment process and the efficiency of managing the elements of the wireless sensor network FANET. Keywords: Wireless sensor network FANET · Estimates of coordinates and derived network elements · Filtering algorithm

1 Introduction Currently, there is an intensive use of the capabilities of single and unmanned aerial vehicles (UAVs) as part of a group in various fields of human activity, including mobile flying FANET sensor networks [1]. To effectively control a group of M UAVs at a ground station, it is not enough to know only their current coordinates in real time, obtained, for example, using onboard GPS sensors. To solve the control problem, information is needed on speeds and accelerations, these parameters depend on many factors and, first of all, on the technical state of the aircraft itself and the state of the atmosphere on its flight path. In addition, in case of loss of control from a ground station, autonomous control from one of the UAVs of the network should be provided. UAV coordinates can be determined using GPS sensors installed on all UAVs and by measuring Dij distances between the i-th and j-th UAVs (Fig. 1), for example, using the IEEE 802.15.4 ZigBee, 6LoWPAN, Thread protocols, RPL [2], in this case, only 3–4 UAVs of the network are equipped with GPS sensors. Information about the motion parameters (speeds, and in some cases, and accelerations of the UAV) can be represented as a state vector (SV) and information about errors in the form of the error correlation matrix (ECM).

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 303–315, 2021. https://doi.org/10.1007/978-3-030-58359-0_17

304

O. Tsukanov and E. Yakornov

Fig. 1. Wireless sensor network of UAVs with GPS sensors

Fig. 2. Rectangular parallelepiped of UAVs coordinate errors

The accuracy of estimating the state of the vector of each UAV network depends on many factors, the main of which are the errors in determining the coordinates of the UAV. Analysis of publications [3–5] suggests that the existing algorithms for estimating the SV do not allow for obtaining high accuracy and at the same time stability of the process of estimating the motion parameters of FANET. The aim of the work is to develop a highly accurate and at the same time stable algorithm for estimating the SV (motion parameters) of FANET elements based on UAV based on information about the coordinates and their errors at a given time.

Estimation of the Motion Parameters of the UAVs FANET

305

2 Formulation of the Problem The mathematical formulation of the problem of estimating the SV of a UAV as FANET elements with their quasi-homogeneous distribution in space is as follows. Based on yi ,  zi , and their rms errors  σ xi ,  σ yi ,  σ zi , it is the available coordinate information  xi ,  − → necessary to determine the motion parameters - state vector estimates Xn coordinates xi yi , zi , derivatives x˙ i , y˙ i , z˙i and ECM element coordinates estimates and further, after the next step of measuring the distances between neighboring UAVs, to refine them. σ yi ,  σ zi , obtained by measuring the EMC determine the coordinates of the UAV  σ xi ,  mutual distances between the UAV network is determined by many factors and depend on the dynamics (speed and acceleration of the flight of the UAV) seasonal, temporal instability of the gravitational field of the Earth, on the propagation conditions of radio waves, etc. 











3 Sustainable Algorithm for Estimating Motion Parameters To obtain highly accurate estimates of the parameters of motion during repeated measurements, it is proposed to use an estimation algorithm based on the Kalman filter [6]. It is assumed that the errors of determining the coordinates, which in turn are input to the algorithm for specifying the location of FANET elements, have the error distribution law, which in general is different from the normal one. For this case, the use of a modified dynamic estimation algorithm based on the Kalman filter allows, in our opinion, to obtain both stable estimates of the aircraft and to ensure their high accuracy. The initial conditions of the estimation algorithm are SV Xn−1 and a priori ECM Kn−1 n-1 measurement step. The state vector can be written as: XTin = |xin , x˙ in , yin , y˙ in , zin , z˙in |.

(1)

The measurement vector is represented as YTin = | xin , y in , z in |,

(2)

whose values are determined by measuring the distance between the UAV. In this case, the extrapolated value of the SV in general can be represented as follows Xine = Fin,n−1 Xin−1 + Gn−1 Un−1 , where Fin,n−1 - type extrapolation matrix:

Fin,n−1

1 t 01 00 = 00 00 00

0 t 1 0 0 0

0 0 t 1 0 0

00 00; 00 00 10 01

(3)

306

O. Tsukanov and E. Yakornov

Gn−1 − control matrix: 0000 0000 0000 = 0 0 0 t 4 0000 0000

Gin,n−1

0 0 0 0 t 4 0

0 0 0 0 0 t 4

  T = Un−1 0 0 0 ax,n−1 ay,n−1 az,n−1  − control vector; t - time between measurements of the distance between the UAV; ax,n−1 , ay,n−1 , az,n−1 − UAV acceleration by coordinates. If to designate Sn−1 − like ECM control vector, then the adjusted value of the a priori ECM control vector can be written as: ∗ Sn−1 = Sn−1 γ ,

(4)

where γ − sustainability parameter of the assessment process. By analogy with (3), the extrapolated a priori estimate of the ECM can also be represented as follows T ∗ T Kine = Fin,n−1 Kin−1 Fin,n−1 + Gn−1 Sn−1 Gn−1 ,

(5)

where Kin−1 – extrapolated prior ECM. Matrix gain can be written as:  −1 Hin = Kine Kine + Qin ,

(6)

where Qin − ECM measurement vectors Yin , na n − m step. Then the estimate of the state vector, presented in the form T    Xin = xin , x˙ in , yin y˙ in , zin , z˙in  ,















can generally be expressed as 

Xin = Xin + Hin (Yin − Cin Xine ),

(7)

where Cin – measurement matrix and aposteriori ECM assessment 

Kin = Kine − Kine Hin , 

wherein Kin expanded form is represented by a matrix of the form

(8)

Estimation of the Motion Parameters of the UAVs FANET

σ x2 sy˙y s Kin = xy sx˙y sxz sx˙z 

sx˙x σ x˙ 2 sx˙ y sx˙ y˙ sx˙ z sx˙ z˙

sxy sx˙ y σ y2 sy˙y syz sy˙z

sx˙y sx˙ y˙ sy˙y σ y˙ 2 sy˙ z sy˙ z˙

sxz sx˙ z syz sy˙ z σ z2 sz˙z

307

sx˙z sx˙ z˙ sy˙z sy˙ z˙ sz˙z . σ z˙ 2

here σ x2 , σ y2 , σ z 2 − error variance by coordinates, σ x˙ 2 , σ y˙ 2 , σ z˙ 2 – velocity error variance, and sxy , syz , sxz – correlation points of error by coordinates, sx˙ y˙ , sy˙ z˙ , sx˙ z˙ – velocity correlation moments, sx˙y , sy˙ z , sx˙z mutual correlation moments of errors in coordinates and velocities. Note that the composition of the measurement vector Yin and state vectors Xine the same, therefore, the measurement matrix will be unity and is not shown in (6)–(7). The above expressions (3)–(8) are the basis of the proposed sustainable vector estimation algorithm. Xin and ECM Kin each UAV. This estimate is made whenever a distance and errors are determined in the FANET UAV network to determine the distance to neighboring UAVs. The hypothesis of UAV motion with a constant value of acceleration between sessions of determining coordinates is taken as a model of movement. In addition, the considered estimation algorithm (3)–(8) makes it possible to ensure a stable estimation of the state vector, including when there is a sudden change in the motion path of the UAV. Estimates of the state of the UAV include coordinates   xin , yin , zin , T . In this case, errors in determining the coordinates σ xi , σ yi , σ zi without taking into account the correlation moments can be graphically represented in the form of a rectangular parallelepiped in the center of which there is a UAV (Fig. 2). Similarly T    - by speed components SV x˙ in , y˙ in , z˙in  , velocity errors by coordinates σ x˙ i , σ y˙ i , σ z˙i (Fig. 2). It can also be represented as a parallelepiped. If the “volume” of such parallelepipeds is small enough, one can speak about the stable auto tracking of each FANET UAV with high precision, and when a high precision auto tracking of the UAV is not required, only its location coordinates are evaluated. The stability of the evaluation of the SV UAV is provided by the “forced increase” of the values of the diagonal elements of the matrix Gin,n−1 σ a˙ x2 , σ a˙ y2 , σ a˙ z2 in (3) by multiplying by γ . This operation, in turn, leads to a lower bound of the values of the elements of the matrix gain K i−1n in (8). Thus, the constants in the algorithm are parameter γ and magnitude t. In this regard, when modeling, you can pre-set ∗ (4) to ensure sustainability and the required the elements of ECM control vector Sn−1 accuracy of estimation. 















4 The Study of the Algorithm for Estimating Motion Parameters The study of this algorithm was carried out by the method of mathematical modeling. For this, a model of UAV movement was used based on the real technical characteristics

308

O. Tsukanov and E. Yakornov

of the Orlan UAV [7], which was described by the 5th degree algebraic polynomial in coordinates: (Fig. 2) 

2 3 4 5 /2 + a3 ti−1 /6 + a4 ti−1 /24 + a5 ti−1 /120, Z t = a0 + a1 ti−1 + a2 ti−1

(9)

4th degree polynomial in speeds: 2 3 4  /2 + a4 ti−1 /6 + a5 ti−1 /24 Z˙ t = a1 + a2 ti−1 + a3 ti−2

(10)

and 3rd degree polynomial on accelerations: 2 3  + a5 ti−1 /6. Z¨ t = a2 + a3 ti−2 + a4 ti−2

(11)

We will conduct a brief analysis of simulation. In particular, the characteristics were studied according to the results of the simulation of the flight of the UAV in the case of straight-line movement of the UAV and the following types of maneuver: avoiding a head-on collision, climb and a smooth turn to 90° with a subsequent transition to straight-line movement. In Figs. 3, 4, 5, and 6 presented the simulation results of the estimation algorithm at the turn on 90° for different parameter values of γ . Here in left picture of Fig. 3 red line shows how the algorithm behaves, and green line shows the trajectory of the UAV in three-dimensional space.

Fig. 3. Simulation results of the estimation algorithm of UAV motion parameters at the turn on 90° for γ = 0.

With γ = 0, (Fig. 3) the algorithm has a divergence of the estimation process, in other words, the well-known Kalman filter is not stable. For 0.4 ≤ γ ≤ 0.8 the proposed algorithm allows to obtain a stable estimation of the motion parameters of the UAV. The results of modeling the operation of the algorithm for estimating the motion parameters during a left UAV rotation of 90 are shown in Fig. 4, 5, and 6. Shows the dependence of the mean square error (MSE) for the XYZ coordinates as function of UAV flight time on the left side, and on the right side is the dependence of the MSE for the derivative of XYZ coordinates (speeds) as function of UAV flight time for γ coefficient value of 0.1, 0.4 and 0.8 accordingly.

Estimation of the Motion Parameters of the UAVs FANET

309

Fig. 4. Simulation results of the estimation algorithm of UAV motion parameters at the turn on 90° for γ = 0.1

Fig. 5. Simulation results of the estimation algorithm of UAV motion parameters at the turn on 90° for γ = 0.4

Fig. 6. Simulation results of the estimation algorithm of UAV motion parameters at the turn on 90° for γ = 0.8

In addition, an analysis was made of the magnitude of the estimation error by coordinate and velocity with averaging over 100 realizations. In straight-line motion, the maximum error in the coordinates is 2 m. − γ = 0.1, and minimal 1.5 m. − γ = 0.8.

310

O. Tsukanov and E. Yakornov

On the maneuver section: maximum error in coordinates 18 m. − γ = 0.1, and minimal 5 m. − γ = 0.4, maximum speed error is 18 m/c. − γ = 0.1, and minimal 10 m/c. − γ = 0.8. The results of modeling the operation of the algorithm for estimating the motion parameters during a left UAV rotation of 90 are shown in Figs. 7, 8, 9, and 10.

Fig. 7. Simulation results of the estimation algorithm of UAV motion parameters avoiding a head-on collision for γ = 0.

Fig. 8. Simulation results of the estimation algorithm of UAV motion parameters avoiding a head-on collision for γ = 0.1

Fig. 9. Simulation results of the estimation algorithm of UAV motion parameters avoiding a head-on collision for γ = 0.4.

Estimation of the Motion Parameters of the UAVs FANET

311

Fig. 10. Simulation results of the estimation algorithm of UAV motion parameters avoiding a head-on collision for γ = 0.8.

In straight-line motion, the maximum error in the coordinates is 4,5 m. − γ = 0.1, and minimal 1.4 m. − γ = 0.8. On the maneuver section: maximum error in coordinates 18 m. − γ = 0.1, and minimal 5 m. − γ = 0.4, maximum speed error is 18 m/c. − γ = 0.1, and minimal 4 m/c. − γ = 0.8. Figures 11, 12, 13, and 14 shows the dependence of the mean square error (MSE) for the XYZ coordinates as function of UAV flight time on the left side, and on the right side is the dependence of the MSE for the derivative of XYZ coordinates (speeds) as function of UAV flight time for γ coefficient value of 0.1, 0.4 and 0.8 accordingly.

Fig. 11. Simulation results of the estimation algorithm of motion parameters climb by UAV for γ = 0.

. In straight-line motion, the maximum error in the coordinates is 4,5 m. − γ = 0.1, and minimal 1.6 m. − γ = 0.8. On the maneuver section: maximum error in coordinates 18 m. − γ = 0.1, and minimal 5 m. − γ = 0.4, maximum speed error is 18 m/c. − γ = 0.1, and minimal 10 m/c. − γ = 0.8. Figures 15, 16, 17, and 18 shows the dependence of the mean square error (MSE) for the XYZ coordinates as function of UAV flight time on the left side, and on the right side is the dependence of the MSE for the derivative of XYZ coordinates (speeds) as function of UAV flight time for γ coefficient value of 0.1, 0.4 and 0.8 accordingly. In addition, an analysis was made of the magnitude of the estimation error by coordinate and velocity with averaging over 100 realizations. In straight-line motion, the

312

O. Tsukanov and E. Yakornov

Fig. 12. Simulation results of the estimation algorithm of motion parameters climb by UAV for γ = 0.1.

Fig. 13. Simulation results of the estimation algorithm of motion parameters climb by UAV for γ = 0.4.

Fig. 14. Simulation results of the estimation algorithm of motion parameters climb by UAV for γ = 0.8.

maximum error in the coordinates is 4,5 m. − γ = 0.1, and minimal 1.6 m. − γ = 0.8. On the maneuver section: maximum error in coordinates 18 m. − γ = 0.1, and minimal 5 m. − γ = 0.4, maximum speed error is 18 m/c. − γ = 0.1, and minimal 10 m/c. − γ = 0.8.

Estimation of the Motion Parameters of the UAVs FANET

313

Fig. 15. Simulation results of the estimation algorithm of motion parameters climb by UAV for γ = 0.

Fig. 16. Simulation results of the estimation algorithm of motion parameters climb by UAV for γ = 0.4.

Fig. 17. Simulation results of the estimation algorithm of motion parameters climb by UAV for γ = 0.4.

In this case, the optimal value of γ with the help of which minimal errors were obtained and the stability of the process of estimating the motion parameters of UAVs as elements of FANET is ensured is 0.8. The minimum value of errors in the coordinates and speeds in the maneuver area is achieved on average in 8 steps of estimation.

314

O. Tsukanov and E. Yakornov

Fig. 18. Simulation results of the estimation algorithm of motion parameters climb by UAV for γ = 0.4.

5 Conclusion The task of estimating the parameters of the UAV aircraft as a FANET network element is solved in two stages: 1) coordinates of the UAV and ECM coordinates are determined; 2) according to the information on coordinates and ECM, obtained using the proposed algorithm on the basis of the Kalman filter, determined values SV and ECM estimation are determined. Simulation results using simulation modeling suggest that the proposed stable estimation algorithm for measuring the distances between UAVs with multiple measurements makes it possible to obtain stable estimates of the coordinate SV and their derivatives and, at the same time, improve the accuracy of determining the coordinates of the FANET UAV, including inaccurate error information distance measurements.

References 1. Romanyuk, V., Stepanenko, E.O., Panchenko, I., Voskolovich, O.I.: Flying self-organizing radio networks. Collection of scientific works of VITI № 1-2017, pp. 105–114 2. Langdon, M.: ZigBee goes underground. E & T Mag., August 2009 3. Zhuk S.Y.: Adaptive filtering of parameters of motion of a maneuvering object in a rectangular coordinate system. In: Zhuk S.Y., Kozhehkurt, V.I., Yuzefovich, V.V. (eds.) Registration, Storage and Processing of Data - 2009, vol. 11, no. 2, pp. 12–24 (2009) 4. Hundunugbo, E.F., Kirichek, R.V., Grishin, I.V., Dumin D.I.: Positioning elements of the sensor network using unmanned aerial vehicles. In: Information Technology and Telecommunications, vol. 2, pp. 26–32 (2016). ISSN: 2307-1303 / RINC. http://sut.ru/doci/nauka/review/20162/2632.pdf 5. Shklyaeva, A.V., Kirichek, R.V., Kucheryavy, A.E.: Test methods for flying sensory networks. Inf. Technol. Telecommun. 4(2), 43–52 (2016)

Estimation of the Motion Parameters of the UAVs FANET

315

6. Grewal, M.S., Andrews, A.P.: Kalman Filtering – Theory and Practice Using MATLAB. Wiley (2001) 7. Melnyk, N.O.: Complete mathematical model of an unmanned aerial vehicle in conditions of motion with disturbing influences. In: Vestnik Voronezh State Technical University, vol. 11, no. 2, pp. 31–33 (2015). ISSN: 1729-6501

Modern Challenges in RadioElectronics Technologies

Universal Complex Model for Estimation the Beam Current Density of High Voltage Glow Discharge Electron Guns Igor Melnyk1(B)

, Sergey Tyhai1(B)

, and Alina Pochynok2(B)

1 Electronic Faculty, Department of Electronic Devices and Systems, National Technical

University of Ukraine “Igor Sikorskiy Kyiv Polytechnical Institute”, Kiev, Ukraine [email protected] 2 Educational and Research Institute of Information Technology, University of the State Fiscal Service of Ukraine, Irpin, Kyiv Region, Ukraine [email protected]

Abstract. The universal complex model to estimation the focal parameters of electron beams, formed by high voltage glow discharge electron guns, is proposed in the article. This complex model consists on the means for calculation of electric field distribution and charged particles trajectories with taking into account the space charge, means for defining the current density distribution in the beam focus on the outlet of electron gun, means for defining plasma boundary form and position, means for defining the beam trajectories in the guiding channel, as well as means for interpolation the beam trajectories in the space of free moving of electrons at the technological chamber. The results of simulation for the distribution of electric field in discharge region as well as for the distribution of current density of formed electron beam on the outlet of gun are given. Also the results for plasma boundary approximation, calculation of electrons trajectories in the guiding equipotential channel, as well as for the distribution of the density of focal beam current on the plane, located in the technological chamber outside the electron gun, are presented in the article. Keywords: Electron beam · Electron gun · High voltage glow discharge · Plasma boundary · Beam guiding channel · Beam trajectories · Current density · Arithmetic-Logic equation · Linear approximation · Square approximation

1 Introduction High Voltage Glow Discharge (HVGD) electron guns are widely used in modern industry for realizing such complex technological operation with of metallic and ceramic items, as welding, brazing, deposition of metallic and ceramics films and coatings, annealing, as well as refining of refractory materials [1–4]. Such guns are operated at the soft vacuum, range of 0.1–10 Pa [5, 6], and have many important advantages relatively to the traditional electron sources with heated cathode, among which such are most valuable and have to be pointed out [7]. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 319–341, 2021. https://doi.org/10.1007/978-3-030-58359-0_18

320

I. Melnyk et al.

1. Operation of electron guns in the soft vacuum lead to simplifying both of gun construction and of evacuation equipment. As a result, the price of electron-beam technological installation is significantly reduced [4]. 2. The HVGD electron guns (HVGDEG) are in the most cases simply disassembled for changing and renewing the spare details, main of its’ usually is a HVGD cathode, cooled by cold water [5–7]. 3. High stability and reliability of HVGDEG operation in the complex physical conditions of industrial electron-beam vacuum equipment, defined by reqirements of realized technological process [5, 6]. 4. Possibility of operation of HVGDEG with different technological gases, including noble and active ones. By this reason choosing of technological gas is mostly defined by realized technological operation, but not by the stability of operation of electron gun and its’ energetic efficiency [5–7]. 5. Simplicity of control the discharge current of both by the aerodynamic way with changing flow of technological gas into HVGDEG [8], and by the electrodynamic way with changing the level of gas ionization in anode plasma by applying the necessary voltage at additional electrode [9]. Taking into account enumerated advantages of HVGDEG, main technological processes, in which its’ can be effectively used as novel advanced instrument, are follows. 1. Welding, brazing and annealing of thin-wall items in electronics production and instrument-making industry. Main advantages of applying HVGDEGs in these technological processes ate cheapness and high productivity of electron-beam equipment and high quality of produced items [5, 6]. 2. Deposition of ceramic thin films in the medium of active gases with possibility of maintaining the chemical reaction between the residual gas and evaporated metal. Main advantages of applying HVGDEGs in this technological process is obtaining of high-quality films with required chemical composition and spatial homogeneity. Stability of mechanical and electrical parameters of obtained ceramic films in repetition production is also provided in this technology. Such technology is very perspective to realizing in mechanical industry, nanoelectronics, as well as in instrument-making industry [1–3]. 3. Pure refining of refractory metals and dielectric materials. Main advantages of applying HVGDEGs in this technological process is cheapness of electron-beam equipment and high purity of obtained ingots. Such technology is very perspective to realizing in modern metallurgy and for refining silicon in electronics production [4]. 4. Realizing of complex technological processes in production of modern multifunctional nanoelectronics components, for example, for obtaining thin ceramics films, used in band high-frequency filters [10–14]. Unfortunately, complex theoretical approaches and suitable approximations for general estimation the current density in the focus of electron beam formed by the HVGDEGs, aren’t existed today. All results, have been presented and analyzed in pervious papers, devoted to simulation of forming the electron beams in HVGD [15] and to its guiding in the equipotential transporting channel [16], were considered separately.

Universal Complex Model for Estimation the Beam Current Density

321

At some pervious investigations on the base of physics of flows of charged particles were analyzed the parameters of electron beams at the focal plane and suitable analytical approximations have been obtained [17, 18]. But the current density distribution in case of propagation of middle-energy beams, range of few or tens kilowatt, in the soft vacuum, wasn’t considered particularly in this works. The energy distribution on the plane, sloped to the beam axis, also wasn’t analyzed. Therefore, the aim of this paper is systemizing of numerical methods for simulation of forming and transporting of electron beam in HVGDEGs and finding the suitable approximation for beam current distribution at the sloped plane.

2 Basic Statement of Considered Problem Basic theoretical presumptions for formulated problem are follows [7–9, 15, 16]. 1. For finding the electric field distribution and the trajectories of charged particles the modified Gomel method have been used. For calculation of electric field for HVGDEGs electrodes’ system geometry finite-different method have been applied [15, 19, 20]. 2. The free moving of electrons in anode plasma is calculated with taking into account Rutherford model of dissipation of electrons on the ions of free gas [17]. 3. Analyzing of plasma boundary form and position have been provided with defining the volume of anode plasma in linear one-dimensional HVGD electrodes system and its recalculation to real geometry of electrodes. Corresponded equations and methodology were considered in papers [5–7, 15]. Verifying the results of calculation of plasma geometry was provided by analyzing the discharge photographs with using methods of image recognizing technique [7, 15]. Suitable mathematical relations for approximations of plasma boundary were also given [7]. 4. Calculation of guiding electron beam, formed by HVGDEGs, in the equipotential channel is provided with taking into account all important physical effects, taking place in the soft vacuum in ionized gases. Distribution of pressure in the guiding channel, which connected the gun and technological chamber, also have been taking into account. Corresponded equations for defining the losses of beam current during its’ transporting were considered in paper [16]. 5. Distribution of current density on the plane, located at the corresponded distance from guns’ outlet and sloped to beam axes, was defined by using the methods of approximation and regression analyze [21]. The basic mathematic relations for realizing the described steps of presented algorithm for iterative calculation the distribution of focal current density of electron beam, formed by HVGDEGs, will be presented in the next sections of this article.

3 Calculation of Electric Field Distribution and the Charged Particles Trajectories Basic equations for calculation distribution of electric field and charged particles trajectories was presented and analyzed at the papers [7, 15].

322

I. Melnyk et al.

The equation for calculation the electric field is based on finite-difference method and generally can be written in the following form [19, 20]:  U n (i, k) = ω Ca U n−1 (i + 1, k) + Cb U n−1 (i, k + 1) + Cc U n (i − 1, k)  ρ n−1 (i, k) n (1) + Cd U (i, k − 1) + Cρ + (1 − ω)U n−1 (i, k). ε0 ρ n (i, k) where i and k – number of current calculated item at longitudinal and radial axis correspondently, n – number of current iterations, ω – relaxation parameter, which usually used for increasing the rate of convergence of iteration process by the potential. For different geometry of simulated electrodes system parameter ω can be changed from 1 till 2 [19, 20]. For providing the calculations at modern computer systems, created for automation of scientific calculations, Eq. (1) for HVGD axially-symmetric systems can be rewritten in the form of arithmetic-logic equation [22]: 1 1 , Dm = 1 − , m = (l > 0) · (l − 1) + (l = 0) · 1, Cm = 1 + 2m 2m ⎛ Uk−1,l +Uk+1,l D U +C U + m k,m h2 m k,l+1 h2r z Uk,l = ⎝(l > 0) · 2 2 + h2r h2z ⎞ Uk−1,l +Uk+1,l 4Uk,l+1 + h2  h2r z ⎠ · Up < Uk,l < Uac + (l = 0) · 2 4 + h2 h2r z

    khr Up − Ua + U = Up · Up + U < Up · + (U ≥ Uac ) · Uac , rp − ra

(2)

where r – radial longitudinal coordinate, z – transversal longitudinal coordinate, hr and hz – discretization steps by coordinates r and z correspondently, k and l – number of the base point by coordinates r and z correspondently, U k,l – calculated value of electric potential, U ac – acceleration voltage, U p – potential of anode plasma, U a – anode potential r p – position of anode plasma at the boundary of considered electrodes system, r a – anode position at the boundary of considered system, m, C m and Dm – additional variables. Obtained graphic dependence for distribution of electric field in HVGDEG electrodes system for acceleration voltage 10 kV is presented at Fig. 1. With knowing electric field distribution trajectories of electrons and ions in simulated electrodes system are calculated with using equation of electron optic, which is written asb follows [17–19, 22]:

 2  ∂U (r,z) ∂U (r,z) dr dr · 1 + − · ∂r ∂z dz dz d 2r  = . (3) m v s s dz 2 2 + 2(U (r, z) − Ub ) where vs – velocity of the charged particle, ms – its’ mass, U b – electric potential on the boundary, which is considered as emitter of charged particles. In HVGDEGs electrodes

Universal Complex Model for Estimation the Beam Current Density

323

Fig. 1. Electric field distribution in the HVGDEG electrodes system, obtained with using Eq. (2).

systems U b = U c for emission of electrons and U b = U p for emission of ions, where U c is the cathode potential and U p is the anode plasma potential. Equation (3) for ions and electrons trajectories is solved numerically with using Runge – Kutt method of four order [20, 23]. The space charge in HVGD gap is defined with using Gomel method, but it was modified for taking into account resonance recharging of ions on the neutral atoms of residual gas. Corresponded formulas for calculation the space charge in the cathode-fall region are written as follow [7, 15]:  Is ρ1s = 2π rt r

ms 2qs



 Ui,k − Ui−1,k 1 1 ; r = r2 − r1; Ur1 = Ui,k + + (lhr − r); Ur1 Ur2 hr

Ni = 2π Na (1 − ξ )rt2 ρi1 =

1−

rlhr Ui,k + Ui−1,k + Ui,k−1 + Ui−1,k−1 f (Uc ); Uc = ; z 4 ρi1 ji1 ; ji1 = ; Ni Ni 1 − 2 rlhr 2 rlhr

2π Na (1−ξ )rt

ρ2s =

z

f (Uc )

Nt

n=1

2π Na (1−ξ )rt

ρ1s ; ρ = ρi − ρe ;

z

f (Uc )

(4)

ρΣ (i, k, i + 1, k + 1) + ρΣ (i, k, i − 1, k − 1) 4 ρΣ (i, k, i − 1, k + 1) + ρΣ (i, k, i + 1, k − 1) + 4 Calculations with using formulas (2–4) are provided iteratively since the relative accuracy for calculated potential in all items of simulation grid at pervious and current iteration became smaller, than established value range of 10−4 [10]. ρi,k =

324

I. Melnyk et al.

In the next part of the article will be considered the iterative algorithm for defining focal beam parameters, based on the methods of minimax analyze [21, 23, 24].

4 Defining of Focal Beam Parameters Lower the main focal beam parameters are focal distance, focal beam radius and beam current density in the focus, from which can be defined the focal power density. Distribution of beam focal current density can be simply defined by interpolation electrons’ trajectories in the region of quasineutral plasma with regarding its dispersion on the ions of residual gas by considering Rutherford scattering model [15]. Focal beam parameters can be obtained by this way with using standard functions of minimax analyze for finding minimum and maximum elements in the defined numerical sets. Corresponded equation system is written at the following form [15]: ⎛

4

⎞

⎛

3

⎞

−4 3 2 2 ⎜ 10 Za ⎟ ⎜ e Za ⎟ min = 2atan⎝ ⎠; max = 2atan⎝ ⎠; v = 2γβ 2 mi v 2 rb (i)

dθ =



4Za (Za +1)re2 γ 2 min n0 dLln  max (γ 2 −1)2



 2eUac , dL = hz 1 − tan2 (ϕ) mi



2 ; dz = Rc − dcp + Rc − dcp − (r0 (k))2 ;

ϕ(i, k) = ϕ(i, k − 1) + d θ;     jbmax (k) = maxi∈i jb (i, k) ; rb (k) = arg 0.7jbmax (k) ; db = 2mink∈k [rb (k)]; Fb = dcp + arg min [rb (k)]hz ,

(5)

k∈k

where i and k – discretization parameters at r and z axis respectively, r b – beam radius at considered plane on coordinate z, jb – beam current density, Rc – cathode sphere radius, d cp – plasma boundary position relatively to the cathode surface, Z – nuclear charge of operation gas atoms, r e – electron radius with respect to Boar model, ϕ – input angle of electron trajectories, θ – the angle of dispersion, γ – relativistic factor. The main difficult in numerical analysis of Eq. (5) is the high convergence angle of HVGDEGs beams, may be greater, then 10° . Furthermore, electron trajectories in anode plasma are usuallty non-paraxial and in this way the beam focus is not briefly-defined. But, in any way, using of minimax algorithm, described in works [16, 19], allow find the focus position F b , as well as beam diameter d b and current density distribution jb (k) at the focal plane. The result of calculation of beam current density distribution for operation pressure 5 Pa and different acceleration voltages are presented at Fig. 2.

Universal Complex Model for Estimation the Beam Current Density

325

jb, A/cm2 1.6 3 4

1.2

2

1

0.8 0.4 1

2

r, mm

3

Fig. 2. Distribution of current density in the focus of HVGDEG electron beam, obtained as result of computer simulation. Acceleration voltage: 1–20 kV; 2–18 kV; 3–15 kV; 4–12 kV.

5 Approximation of Plasma Boundary Geometry and Its Position Relatively to Cathode Lower As have been pointed out early, for used in industry electrodes system with spherical cathode and hollow anode [7, 10] plasma boundary for large values of discharge current is usually considered as the electrode with fixed potential and as a source of ions. For real electrodes systems plasma boundary can be defined by analyses the brightness of discharge photographs, which obtained experimentally and treated at computer with using of image recognizing technique [7]. On such photographs’ plasma boundary is corresponded to boundary line between the light and dark regions in longitudinal direction from cathode to anode. The basic theoretical presumption for defining the plasma boundary position is that for large value of beam current, greater than 200 mA, position of plasma boundary is stabilized and its geometry is similar to the cathode surface [5, 6]. Taking into account this fact, firstly plasma boundary position calculated the for one-dimensional HVGD system with the similar volume, and after that in recalculated for the real electrodes geometry. Corresponded equation for one-dimensional planar axially-symmetric HVGD system can be written as follows [5, 6]: l =l− dcp

Id  mi Qi0 χγa + me −

5μeo kTe 2 e R2 pa0



mi kTi

,

(6)

where l and R – longitude and transversal sizes of planar electrodes system respectively, pa0 – residual pressure in discharge region, I d – discharge current, Qi0 – average ions recharge cross-section, χ – average coefficient of elongation of electrons’ trajectories in anode plasma, γa – coefficient of electrons reflection from anode surface, me – electron mass, mi – mass of gas ions, μe0 – mobility of electrons in plasms, e – charge of electron, k – Boltzmann constant, Ra – radius of anode cone founding, ha – anode high, and la – anode cone generator.

326

I. Melnyk et al.

Taking into account, that plasma volume in system with conic anode is the same, as in linear cylindrical system, corresponded equations for defining the plasma boundary position in the electrodes system with spherical cathode and conic anode can be written as follows [7]:       4 l 2 8  3 2 l4d l l  9Ra dcp la 12 6  3R 2R8a la16 dcp la Ra la24 R12 a a cp  a + − − , dpa =  4 2 − 6 6 4 4 3 2 2ha sw 4h8a s2 w2 243h30 9h19 27h15 a s w a s w a s w s · h2a , dcp = ha − dpa , + R2a   2  2 Ra Ra − 1+ , s=1+ ha ha

la2 = h2a + R2a , w = 3 −

h2a

(7)

where ha – transversal sizes of HVGD electrodes system, corresponded to highness of anode surface, Ra – radius of basis of anode surface, la – cone forming line of the anode surface, d cp – plasma boundary position relatively to the cathode surface, w and s – additional variables. For expert, who designed the HVGDEG, dependence of plasma boundary position relatively to the cathode d cp verse discharge current I d d cp (I d ) is usually very important. Corresponded dependences, obtained for different operation pressure pa0 for acceleration voltage U ac = 15 kV for geometry parameters ha = 0,08 m and Ra = 0.035 m with using formulas (6, 7) are presented at Fig. 3. For using nitrogen as operation gas, aluminum as a cathode material and cooper as anode material, such internal parameters of HVGD m2 lighting was choose: Qi0 = 5.3 · 10−19 m−2 , μi0 = 1.27 · 10−4 V·s , Te = 800 ◦ K, γa = 4.6, χ = 1.2 [5, 6]. It should be pointed out, that the same dependences for function d cp (I p ) was obtained experimentally, disagreement between theoretical and experimental data was in range 10–15% [15]. Since for small values of discharge current I d geometry of plasma boundary isn’t correspond to the form of the cathode surface, separate investigation to defining the plasma boundary curve from the discharge photographs with using image recognizing technique are necessary [7, 15]. The methodology of analyzing the discharge photographs is simple. One can see that corresponded to Fig. 1 the distance from the cathode to anode hole d ca and to the anode diaphragm d cd , are the known constructive parameters of simulated and investigated HVGD electrodes’ system, the basic points of plasma boundary are defined as the region of increasing the brightness of photograph with changing the coordinate z. Corresponded equation is written as follows: ph

r dcp

=

ddp

dr ph ca

dca

r + dcd ,

(8)

where d dp – distance from diaphragm to plasma boundary, at the photograph, d ca – distance between cathode and anode hole, d dp – distance from cathode to diaphragm. All upper indexes r in Eq. (8) corresponded to real dimension and indexes ph – to dimensions at photograph.

Universal Complex Model for Estimation the Beam Current Density

327

Fig. 3. Dependences of cathode-plasma distance d cp verse discharge current I d for different operation pressure.

One of analyzed photograph, obtained for simulated electrodes system for accelerated voltage U ac = 10 kV and operation pressure p = 1 Pa, is presented at Fig. 4. The problem of using Eq. (8) is that the changing of brightness on transversal coordinate is smooth, and by this reason direct defining the correct point, where it is changed, is impossible. Therefore, firstly defined the region of brightness changing, and after that the mean point in defined range of coordinate z considered as correct. In the theory of regression analyze such methods is well-known as method of slide mean value [21]. After defining the set of basic point of plasma boundary its approximation provided with using analytical function f (r), which was choose by complex investigation of behavior of this function for different discharge regimes. The followed function is written as:  2 r + B, (9) z = f (r) = A exp − α where A, B and α – empirical coefficients. These coefficients are defined for the basic sets of points (r i , zi ), i = 1, …, n, where i – point number in the set, n – total amount of points, with using well-known least squares method, described mathematically by following minimax relation [15, 16, 18, 19]:

 n  F(A, B, α) = min .. (10) (f (ri ) − zi )2  i=1

A,B,α

I. Melnyk et al.

r, m

328

z, mm Fig. 4. Digital photograph of HVGD gap for analyzing plasma boundary geometry.

Fig. 5. Approximation of plasma boundary geometry with using Eq. (9).

Finding the minimum of function F(A, B, α) for different coefficients A, B and α with using Eqs. (9, 10) is the special complex problem of regression analyze and it solved numerically with using the simplex Needler-Mead method [21, 22, 24]. Approximation results were obtained in MatLab computer system and both symbolic processor and numerical functions of this system are used complexly [21]. The result of approximation the plasma boundary geometry for photograph, presented at Fig. 4, is given at Fig. 5. Corresponded parameters for approximation function (9) are: A = 0,0039 m, B = 0,0485 m and α = 0,0281 m2 .

6 Simulation of Beam Guiding in the Equipotential Channel Since the required operation pressure in the gun and technological chambers for such complex technological operations, as deposition of ceramic coatings and refining of refractory metals may be different [1–4, 7, 8], guiding of electron beam in the equipotential channel is used in technological equipment for maintaining the required difference of operation pressure. Corresponded scheme of electron beam vacuum equipment is presented at the work [8]. The main task of simulation the propagation of electron beam in an equipotential channel is also the problem of minimax analyze, and it formulated as minimizing the losses of beam current in channel with providing the necessary pressure difference between gun and technological chamber [16]. Main presumptions of this model are follows. 1. Beam guiding in the cylindric and conic pipes is considered. It was proved also by numerical calculations, that using of the pipes with more complex geometry, defined by the power function r(z) = (z – a)α , isn’t effective, since combination of input diaphragm with cylindric and conic pipes allow obtain the difference of pressure range of 10−2 with relative beam losses in channel smaller, than 0,01% [16]. 2. For reducing the beam losses focusing magnetic lenses are used, and special location of lenses, as well as value of magnetic field are also optimized for reducing the pressure.

Universal Complex Model for Estimation the Beam Current Density

329

For forming the complex model of beam guiding system such approaches have been used [16, 22]. 1. For defining the divergence angle of electron beam the model of analyzing the trajectories of electrons in the anode plasma region, described in Sect. 4 of this article, have been used. Corresponded equation for finding this angle ϕ is written as follows:   mink∈k [rb (k)] − mink∈k [rb (k − 1)] ϕ = atan . (11) hz 2. For calculation the pressure difference between the gun and technological chamber the Knudsen model for molecular regime of gas flow have been used. For cylindrical pipe with diaphragm such set of equations is usually correct [22]: pc =

pg Ucyl R2 − R2 lp 1 ; Wcyl = 2 2 1 2 + ; Ucyl = , Ucyl + Sp , Wcyl 116π R2 R1 968R32

(12)

where S p – productivity of vacuum pumping system, pg – pressure in the guns’ chamber, pc – pressure in the technological chamber, R2 – the channel radius, R1 – the diaphragm radius, lp – length of channel, Wcyl – vacuum resistance of cylindrical guiding channel with diaphragm, Ucyl – its’ conductivity,  pg Ucon pc = , Ucon + Sp ,

Ucon =

4R1 R2

8R0 T πM

3lp (R1 + R2 )

,

(13)

where R0 the universal gas constant, T – the temperature of operation gas with taking into account it heating by the beam electrons, M 0 – the molecular mass of operation gas. For relatively high pressure in discharge chamber of electron gun intermediate gas flow regime in the guiding channel is established. In this case for calculation channel conductivity using of correction coefficient is necessary [16]: J =

1 + 202(R1 + R2 )¯p + 2653((R1 + R2 )¯p)2 ; Uim = JUm , 1 + 236(R1 + R2 )

(14)

where p¯ – average gas pressure in the guiding channel, U m – value of conductivity for m 0, 5 ≤ Dl /Sl ≤ 2,

(15)

where S l – wideness of nonmagnetic lens gap and Dl – thickness of winding wire region. At the conditions, defined by relation (15), axial magnetic field Bz0 for short focusing lens with ferromagnetic casing with accuracy nearly 5–10% defined as [16, 17]: ⎛ ⎞ Bz0 =

zl + S2l 1, 257 · 10−4 Il Nl ⎜ ⎜ ⎝ 2 2Sl Dl + zl + 3

Sl 2

2 − 

zl − S2l

2 Dl + zl − 3

Sl 2

⎟ ⎟

2 ⎠, (16)

330

I. Melnyk et al.

where Il – lens current, Nl – number of lens coils and zl – wideness of lens. olecular regime, U im – corresponded value of conductivity for intermediate regime and J – the semiempirical coefficient for recalculation of vacuum conductivity for different regimes. 3. For calculation axial magnetic field of short focusing lenses were applied the approximation, which give the suitable precision, when the corresponded inequality is came to true [16]: With known, from relation (16), the axial magnetic field distribution Bz0 (z), its distribution for any point of guiding channel is simply obtained as Taylor series expansion for Br and Bz components. Corresponded relations are written as follows [16, 17]: r  r 3  r 2  Br = − Bz0 + Bz0 + O r 3 , Bz = Bz0 − Bz0 + O r 3 . 2 16 4

(17)

Values of input angle, pressure and magnetic induction, obtained with using relations (11–17), presented in the iterative steps 1–3 of considered calculation algorithm, were used on the next steps of simulation for calculation electrons’ trajectories in guiding channel. 4. Calculation of beam boundary trajectory at the current iteration. For providing this step of iteration procedure the complex set of algebraic and differential equation have been used, since for simulation the propagation of short-focus electron beam in the middle acuum a lot of physical effects, which are greatly influence to beam trajectories. Among these effects most important are follows [16]. – Similar to simulation of free moving of beam electrons in anode plasma, which have been considered at Sect. 4, Rutherford scattering model for dissipation of beam electrons is considered. – Influence of ions’ focusing into trajectory of beam electrons, which is strongly depended on pressure of operation gas in the considered part of guiding channel. – Overcompensation of the space charge of electron beam by the ions of residual gas. – Magnetic gas focusing as relativistic effect. Calculation results are shown that for middle energies of beam electrons range of 5–30 keV, which corresponded to HVGD lighting [5–7], influence of ion overcompensation and magnetic gas focusing isn’t very significant, smaller than 1% relatively to influence of scattering and ions’ focusing [16, 22]. On the ground of theory of electron beams propagation this fact can be simply explained by the relatively large diameter of transported beam [17, 18].

Universal Complex Model for Estimation the Beam Current Density

331

Therefore, the complete set of equation for defining the beam boundary trajectories in the guiding channel is written as follows [16]: ⎛ 3 ⎞

 4 2 10−4 Z 3 ⎝ Za ⎠; = 2atan min = 2atan ;  max 2γβ2 2γβ2 

 M ε0 ne Uc d 2 rb ni0 = exp − dz + θs ; ; θ = me Uc dz 2 ε0 ne rb2  2 Ib 1 − f − β 2 erb Bz0 d 2 rb C ne  ;C = ; = − , f = 3/2 ni0 − ne dz 2 rb 8me Uac 4π ε0 m2ee Uc √

π rb2 Bi pne

(18)

where Za – charge of nuclear for residual gas atoms, β = ve /c – relation of electron velocity ve to light velocity c, θ – corresponded slope angle of beam electrons, θmin and θmax – minimal and maximal scattering angles correspondently, θs – changing of slope angle of beam electrons by its scattering on the atoms of residual gas, θ – average angle of electrons dissipaion, rb – radius of guiding beam on the corresponded part of channel, ni – concentration of ions on the corresponded part of guiding channel, Bi – level of ionizing, p – gas pressure on the corresponded part of guiding channel, ne – concentration of electrons, ε0 – dielectric constant, f – level of space charge compensation by ions of residual gas and I b – current of electron beam, which on this step of algorithm is considered as stable value. 5. Model of losses of beam current during it guiding in the pipe is described by the following equations [16]:

j(r) = j0 exp −  dIb =

πβb2 j0

rb2 βb2

 , Ibn = Ibn−1 − dIbn ,

(19)

      rb 2 r b − βb 2 exp − − exp − , 2 2

where j(r) – function of current density distribution, j0 – axial current density, βb – parameter of distribution function, characterized dissipation of electrons velocities, n – number of iteraion. Iterations for steps of Algorithm 2–5 with using Eqs. (11–19) were realized among the whole length of channel with specified discretization step on coordinate z hz . During providing these calculations in Eqs. (12, 13) assumed, that l p = z, and by this way distribution of the pressure among the channel is defined [16]. Obtained simulation results are presented in Fig. 6. The parameters of guiding system, have been used for providing these calculations, are presented in Table 1.

332

I. Melnyk et al.

Fig. 6. Dependence for beam trajectories (upper) and losses of beam current (lower), obtained with using Eqs. (11–19) for guiding system parameters, presented at the Table 1. Calculated data were obtained for different inlet angle θ. Table 1. Parameters of simulated beam guiding system. Parameter

Value

Parameter

Value

Acceleration voltage, U ac

104 V

Diameter of channel pipe, D2 = 2R2

25 mm

Starting current of electron beam, I b 1 A

Pressure in HVGDEG, pg

5 Pa

Starting radius of electron beam, r b 3 mm

Pressure in technological chamber, pc

10−1 Pa

Diameter of channel input hole, D1 = 2R1

9 mm

Productivity of pumping system, S p 0.1 m3 /s

Number of coils of magnetic lens, Nl

2000

Current of magnetic lens, I l

2.5 A

Wideness of lens, zl

50 mm

Wideness of nonmagnetic lens gap, Sl

10 mm

Parameters of focusing magnetic lens

7 Interpolation of Beam Trajectories Outside the Electron Gun The basic assumption for formulation the task of interpolation the beam boundary trajectory in technological chamber is minimizing the amount of basic points, which should to be used. Since, as its clear form Fig. 6 and experimental results, the beam boundary

Universal Complex Model for Estimation the Beam Current Density

333

trajectory in the free space is mostly linear, using of 3 points is enough for suitable interpolation with high accuracy [17, 18, 21]. With using such approach corresponded relation have to be satisfied: h1 < hf < h3 ,

(20)

where h1 and h3 – left and right basic points respectively, hf – focal radius of electron beam, which corresponded to minimal value of r(h) dependence. The main idea of this theoretical presumption for linear and square interpolation is clear from Fig. 7. At this figure basic points are noted as (h1 , r 1 ), (h2 , r 2 ) and (h3 , r 3 ), (hfs , r fs ) – coordinates of beam focus for square interpolation and (hfl , r fl ) – coordinates of beam focus for linear approximation.

r 2

r1

1

r3 r2

rfs rfl

h1

hfl

h2 hfs

h3

h

Fig. 7. Example of linear (1) and square (2) interpolation with using three basic points

Corresponded equation for linear interpolation in simple form with using the absolute value function can be written as follows:    rl (h) = k h − hfl  + rfl . (21) were the coefficient k is simply defined as: k=

r1 − r2 . h1 − h2

(22)

The position of beam focus by the coordinate h for linear interpolation function (21) is defined from relation:   1 r3 − r2 , (23) hfl = · h3 + h2 + 2 k and corresponded beam radius, which used as parameter in Eq. (21), is defined from the coordinates of basic points as: rfl =

kh2 + 3 · (r3 + k · h3 ) − r2 . 2

(24)

334

I. Melnyk et al.

The Eq. (21) is simple and generally it can be used for describing the boundary trajectory of electron beam, but the disadvantages of such interpolation are follows [17, 18]. 1. The derivation of function (7) in the point with coordinate (hfs , r fs ) isn’t existed. 2. The focal beam diameter r fs , obtained with using formulas (6), usually have lowest value than real. For finding more correct interpolation function to describing the correct geometry of boundary trajectory of electron beam, such iterative algorithm has been used. 1. Form the selected set of interpolation points (h1 , r 1 ), (h2 , r 2 ) and (h3 , r 3 ) choosing one, for which the value of r c is minimal, namely [20, 21, 23, 24]: rc = min(r1 , r2 , r3 ).

(25)

2. Form the point with coordinate (hc , r c ), found with using Eq. (25), have been formed the corresponded square approximation function: p(h) = Cc2 h + Cc1 h + Cc0 ,

(26)

where the coefficients C c2 , C c1 and C c0 obtained with using following formulas:

 k hc , Cc1 = k · 1 −  , Cc0 = rc − Cc2 h2c − Cc hc . (27) Cc2 =  2 hc − hfl 2 hc − hfl With using such approach approximation function for beam boundary trajectory is written by following arithmetic-logic equation, which included functions (21), (26) and (27):    rc (h) = (h > hc ) ∨ h < 2hfl − hc · p(h)    + (h ≤ hc ) ∨ h ≥ 2hfl − hc · rl2 (h) (28) Basic conceptions of forming the arithmetic-logic equations and it’s using in numerical calculation algorithm was considered in the paper [22]. It was also pointed out in this paper, that such approach is most effective with applying corresponded methods of matrix programming and can be effectively realized in parallel algorithms. Equation (28) can be considered as combined liner-square approximation of boundary trajectories of electron beams, formed by the HVGDEGs. The results of interpolation beam boundary trajectory with using Eqs. (21–28) were considered for different sets of points, presented in Table 2, and corresponded graphic dependence are presented at Fig. 8. It is clear from obtained graphic dependences, presented at Fig. 8, that focal beam radius r fc for equation of combined interpolation (28) is usually grate than the value r fl , defined by Eq. (24). But, in any case, obtained Eq. (23) for beam focus position is correct. Therefore, the corrected value of focal beam radius rfc is simply defined by substitution

Universal Complex Model for Estimation the Beam Current Density

335

Table 2. Three considered sets of points for testing the arithmetic-logic interpolation function (28) Number of set (h1 [mm], (h2 [mm], (h3 [mm], r 1 [mm]) r 2 [mm]) r 3 [mm]) 1

(0.2, 1.3)

(0.3, 1.1) (0.6, 1.5)

2

(0.2, 1.5)

(0.4, 1.3) (0.5, 1.4)

3

(0.2, 1.5)

(0.4, 1.2) (0.5, 1.3)

Fig. 8. The results of interpolation of electron beam trajectories for three sets of data, presented at Table 2

the value hfl , obtained from the set of basic points with using Eq. (23), into the relation of square interpolation (26). Corresponded analytical relation for calculation the beam radius rfc is written as follows: rfc = Cc2 hfl + Cc1 hfl + Cc0 , where the coefficients C c2 , C c1 and C c0 are defined by the set of Eq. (27).

(29)

336

I. Melnyk et al.

8 Calculation of Beam Current Density Distribution at the Plane Surface, Sloped to Beam Axis It is well-known fact that at HVGD the current density of formed electron beam is defined by Gauss distribution [5, 6]:

 r2 (30) j(r) = j0 exp − 2 , βb where j0 – maximal current density, which corresponded to condition r = 0, βb – parameter of distribution, defined as:  Ib βb = , (31) 2π j0 where I b – the beam current. Gaussian distribution of beam current density in HVGD is caused by multiple interactions of electrons with the atoms and ions of residual gas, therefore the main low of multiple factor analyzes, namely, Gaussian law, is strongly fulfilled in this case [24]. And Eq. (31) for βb is the result of transmission from uniform to Gaussian distribution, it has been obtained in the paper [21]. Clear, that the set of Eqs. (30, 31) is non-linear and have to be solved numerically. The problem of its solving is simplified, if we taking at the first iteration step the initial value of current density as: j0 =

Ib . 2 2π rfc

(32)

In Eq. (32) parameter rfc is defined by relation (29), considered in the previous section. With using approach, defined by Eq. (32), the set of Eqs. (30, 31) can be rewritten as one non-linear equation relatively to variable j0 :

 2 2π j0 rfc Ib = j0 exp − . (33) 2 Ib 2π rfc Equation (33) is solved numerically with using Stephenson method [23]. For realizing this solution, the relation (33) firstly has to be rewritten at the canonic form:

 2 2π j0 rfc Ib = 0, (34) f (j0 ) = j0 exp − − 2 Ib 2π rfc and after that corresponded iteration formula of Stephenson method have been applied [23]:

Universal Complex Model for Estimation the Beam Current Density

 f 2 (j0 )n Ib  ,  , (j0 )n+1 = (j0 )n −  (j0 )1 = 2π r f (j0 )n + f (j0 )n − f (j0 )n

337

(35)

where n – iteration number, or the considered step of Stephenson algorithm. The main task of provided theoretical investigations is defining of electron beam current density at the sloped plane. Corresponded geometrical model is presented at Fig. 9.

h δ

Electron beam Sloped plane

α hfl

r

h

rfc Fig. 9. The geometrical model of irradiation the sloped plane by the electron beam.

It is clear from Fig. 9, that in considered geometrical model α is the angle of sloping of the plane, located under the beam action, and geometrical parameters hfl and r fc defined by relations (23, 29), have been considered in the previous section. Corresponding to the Fig. 9, the highness of small piece of plane, located strongly under the beam, is calculated as follows: ha = hfl + δh = hfl + r tan(α).

(36)

With known value of highness h corresponded value of beam radius calculated by Eq. (29). After that with using iterative Eq. (35) obtained the current density, which corresponded to current coordinates (h, r). For defining the correct value ha Eq. (36) have been used. The results of calculation of current density distribution for different values of item position h and slope angle α are presented at Fig. 10. All results have been obtained for beam current 10 A. Dependences of position of the point of maximal current on the angle of plane slope for different values of its position are presented at Fig. 11. Such dependences are very important to defining the best technological regimes during elaboration the electron beam installations, based on high voltage glow discharge electron guns.

338

I. Melnyk et al.

Fig. 10. Distribution of current density at the sloped plane for different angle α and highness h: a – set #1, h = 0,35 mm; b – set #1, α = 5o .

Universal Complex Model for Estimation the Beam Current Density

339

Fig. 11. Dependences the position of the point of maximal current on the angle of plane slope for its different position h

9 Conclusion Universal complex approach to simulation of high voltage glow discharge electron guns, considered in this article, included the follows necessary steps. 1. Defining the plasma boundary position with using Eqs. (6, 7). For more precision defining the geometry of plasma boundary computer analyze of discharge photograph with using relation (8) and approximation relation (9) have been used. 2. Calculation the distribution of electric field in considered HVGD electrodes system with using arithmetic-logic Eq. (2). 3. Calculation of the trajectories of electrons and ions by solving the differential Eq. (3) with using Runge – Kutt method of four order. 4. Defining the space charge in considered HVGD electrodes system with using set of Eq. (4). 5. Recalculation of electric field and trajectories of charged particles by the steps 2 and 3 of this algorithm with taking into account the space charge, defined at step 4. Steps 2–4 is fulfilled iteratively till obtaining the required accuracy of calculations by the electric potential. 6. Defining the focus position and focal beam diameter on the outlet of HVGDEG with using set of Eq. (5). 7. Defining the optimal geometry of beam transporting channel with using the set of Eqs. (11–19).

340

I. Melnyk et al.

8. Interpolation of boundary trajectory of electrons in the technological chamber and defining the focus position hfl and focal beam diameter rfc with using the set of Eqs. (20–29). 9. Defining the beam current distribution at the sloped plane with using the set of Eqs. (30–36). Using of arithmetic-logic equations at the steps 2–4, 7 and 8 allow to optimize the software realization of proposed methodology of calculations with using the novel methods of logic and matrix programming. Corresponded theory of matrix programming, as well as analyzing the possibility of realizing such approach at the seventh step of described methodology, have been briefly considered in the paper [16]. Proposed complex methodology for simulation the action of electron beam, formed by the glow discharge electron guns, to the treated item, sloped to beam axes at specified angle, give the excellent possibility to estimate the effectivity of industrial using of such advanced technologies. Obtained simulation results are generally corresponded to experimental data, obtained early in the Laboratory of Electron Beam Technological Equipment, Faculty of Electronic, National Technical University of Ukraine. Some examples of comparison the theoretical and experimental results were presented early in papers [7, 8, 15, 16]. Proposed methodology of simulation, as well as presented in the article simulation results, are interesting and important to wide range of qualified experts in the branch of simulation and elaboration of novel types of industrial electron sources. Also, it’s should be interesting to experts, who elaborated the novel industrial technological electron-beam equipment and put it into operation.

References 1. Feinaeugle, P., Mattausch, G., Schmidt, S., Roegner, F.H.: A new generation of plasma-based electron beam sources with high power density as a novel tool for high-rate PVD. In: Society of Vacuum Coaters, 54th Annual Technical Conference Proceedings, Chicago, pp. 202–209 (2011) 2. Mattausch, G., et al.: Gas discharge electron sources – proven and novel tools for thin-film technologies. Elektrotech. Electron. 49(5–6), 183–195 (2014) 3. Grechanyuk, M.I., Melnyk, A.G., Grechanyuk, I.M., et al.: Modern electron beam technologies and equipment for melting and physical vapor deposition of different materials. Elektrotech. Electron. (E + E) 49(5–6), 115–121 (2014) 4. Ladokhin, S.V., et al.: Electron Beam Melting in the Foundry Production. Stal, Kyiv (2007). (in Russian) 5. Novikov, A.A.: High Voltage Glow Discharge Electron Sources With Anode Plasma. Energoatomizdat, Moscow (1983). (in Russian) 6. Zavialov, M.A., Keindel, Yu.E., Novikov, A.A., Shanturin, L.P.: Plasma Processes in Technological Electon Guns. Atomizdat, Moscow (1989). (in Russian) 7. Denbnovetskiy, S., et al.: Principles of operation of high voltage glow discharge electron guns and particularities of its technological application. In: Proceedings of SPIE, The International Society of Optical Engineering, pp. 10445–10455 (2017)

Universal Complex Model for Estimation the Beam Current Density

341

8. Denbnovetsky, S.V., Melnyk, I.V., Melnyk, V.G., Tugai, B.A, Tuhai, S.B.: Simulation of dependences of discharge current of high voltage glow discharge electron guns from parameters of electromagnetic valve. In: 2017 IEEE 37th International Conference on Electronics and Nanotechnology (ELNANO). Conference Proceedings, Kyiv, Ukraine, April 18–20, 2017, pp. 369–374 (2017) 9. Melnyk, I.V.: Simulation of energatic efficiency of triode high voltage glow discharge electron sources account of temperature of electrons and its’ mobility in anode plasma. Radioelectron. Commun. Syst. 60, 413–424 (2017) 10. Zakharov, A., Rozenko, S., Ilchenko, M.: Two types of trisection bandpass filters with mixed cross-coupling. IEEE Microw. Wirel. Compon. Lett. 28(7), 585–587 (2018) 11. Zakharov, A., Rozenko, S., Litvintsev, S., Ilchenko, M.: Trisection bandpass filter with mixed cross-coupling and different paths for signal propagation. IEEE Microw. Wirel. Compon. Lett. 30(1), 12–15 (2020) 12. Zakharov, A., Litvintsev, S., Ilchenko, M.: Trisection bandpass filters with all mixed couplings. IEEE Microw. Wirel. Compon. Lett. 29(9), 592–594 (2019) 13. Zakharov, A., Rozenko, S., Ilchenko, M.: Varactor-tuned microstrip bandpass filter with loop hairpin and combline resonators. IEEE Trans. Circuits Syst. II Exp. Briefs 66(6), 953–957 (2019) 14. Zakharov, A., Litvintsev, S., Ilchenko, M.: Transmission line tunable resonators with intersecting resonance regions. IEEE Trans. Circuits Syst. II Exp. Briefs 67(4), 660–664 (2020) 15. Melnyk, I.V.: Numerical simulation of distribution of electric field and trajectories of the particles at electron sources, based on high voltage glow discharge. Radioelectron. Commun. Syst. 48(6), 61–71 (2005) 16. Denbnovetsky, S.V., Melnik„ V.G., Melnik, I.V., Tugay, S.B.: Simulation of guiding of shortfocus electron beams from low to high vacuum with taking into account dessipation of electrons thermal velosity. In: Proceedings of SPIE, Seventh Seminar on Problems of Theoretical and Applied Electron and Ion Optics, vol. 6278, pp. 627809-1–627809-13 17. Szilagyi, M.: Electron and Ion Optics. Springer Science & Business Media (2012) 18. Lawson, J.D.: The Physics of Charged-Particle Beams. Clarendon Press, Oxford (1977) 19. Hockney, R.W., Eastwood, J.W.: Computer Simulation Using Particles. CRC Press (1988) 20. Mathews, J.H., Fink, K.D.: Numerical Methods. Using Matlab. 3rd edn. Amazon (1998) 21. Draper, N., Smith, H.: Applied Regression Analysis, 3rd edn. Wiley Series (1998) 22. Luntovskyy, A.O., Melnyk, I.V.: Simulation of technological electron sources with use of parallel computing methods. In: XXXV IEEE International Scientific Conference “Electronic And Nanotechnology (ELNANO)”. Conference Proceedings, Kyiv, Ukraine, April 21–24, 2015, pp. 454–460 (2015) 23. Samarskiy, A.A., Gulin, A.V.: Chislennye Metody [Numerical Methods], 432 p. Science, Moscow (1989). (in Russian) 24. Vasiliev, V.P.: Chislennye metody resheniya ekstremal’nyh zadach. Uchebnoe posobie dlya vuzov [Numerical Methods for Solving of Extremely Problems. Tutorial Book for Institutes of Higher Education]. Nauka, Moscow (1988). (in Russian)

The Radiation Characteristics of Two Coupled Vertical Dipoles with a Finite Size Screen Nadezhda Yeliseyeva(B)

, Sergey Berdnik(B)

, and Victor Katrich(B)

V.N. Karazin, Kharkiv National University, Kharkiv, Ukraine [email protected]

Abstract. The 3D vector problem of diffraction of the fields of two coupled vertical electric dipoles placed over an infinitely thin rectangular ideally conducting screen is considered. On the base of the asymptotic solution to this problem by the method of the uniform geometric diffraction theory, fast algorithms and software are developed for computation of the radiation patterns and directive gains at maximum radiation, as well as radiation resistances as functions of the electric dimensions of the radiating system. It is shown that, when the removal of dipoles from screen is fixed, the appropriately chosen distance between the dipoles and the appropriately chosen dimensions of the screen provide for the symmetric patterns with high directive gain. When the dipoles currents are in quadrature, the patterns may be unidirectional. For verification of the results, the patterns and directive gains of the radiating systems are calculated using Feko software. Keywords: Rectangular screen · Electric dipole · Diffraction · Edge wave · Radiation pattern · Field amplitude · Directive gain · Radiation resistance

1 Introduction The dipole antennas being the most widely used in antenna practice, in VHF band are often mounted near to metal bodies with various configurations [1]. Rather often the antennas are used with a reflector in the form of a rectangular plane screen. Therefore, it is important and actual for practical purposes to understand the physics of the formation of radiation patterns (RPs) of a dipole antenna, depending on the screen sizes, the screen side ratio at different dipole orientation relative to the screen. It is known that in the case of a dipole parallel to the screen, changing the screen side’s dimensions and their ratio makes it possible to control the Back-to Front Ratio of antenna, the level of radiation in a given direction [2]. In the case of two dipoles with a screen, additional opportunities appear to influence the patterns due to changing distances between the dipoles and different phase differences between their currents. Within the framework of the uniform geometric theory of diffraction (UGTD), such studies were carried out in [3] for two half-wave dipoles located parallel to a square screen at a distance from the screen equaled to the wavelength quarter. It is shown that the appropriately chosen distance between the in-phase dipoles and the appropriately chosen dimensions of the screen provide for © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 342–358, 2021. https://doi.org/10.1007/978-3-030-58359-0_19

The Radiation Characteristics of Two Coupled Vertical Dipoles

343

axially symmetric patterns in the main observation planes and the maximum maximorum directive gain due to the influence of edge effects. In [4], an asymptotic solution of the 3D problem of diffraction of a field radiated by dipole situated perpendicularly above a rectangular screen, depending on the sizes and shape of the screen, was obtained and analyzed. It is proved that in this case the ratio of the screen sides affects the patterns in the main meridional observation plane and does not affect in the azimuth plane. The direction and intensity of the maximum radiation also depends on the size and shape of the screen. Now the 3D vector problem of diffraction of the fields of two coupled electric dipoles placed in perpendicularly over an infinitely thin rectangular ideally conducting screen is solved. The fast-active algorithm and software for computing the patterns [5], as well as the directive gain and radiation resistance are developed, and the physical causes of obtained results are analyzed.

2 Statement of the Problem Consider a radiating structure consisting of two identical symmetric electric dipoles 1 and 2 with arm length l located perpendicularly to perfectly conducting rectangular infinitely thin screen with side dimensions L and W (Fig. 1).

Fig. 1. Geometry of the problem.

Let us introduce two interrelated the Cartesian coordinate systems XYZ and X Y Z, and two spherical frames (R, θ, ϕ) and (R, θ , ϕ ) with the origin in the screen midpoint O. The distance between the dipoles along the Y-axis is ξ = 2d, and the distance between the screen and dipoles along the X-axis is h. In addition, we introduce Cartesian coordinate

344

N. Yeliseyeva et al.

systems X 1 Y 1 Z 1 and X 2 Y 2 Z 2 with the origins at points O1 and O2 that are projections of the dipoles’ centers on the screen. Within the frame of UGTD method the structure of full electrical field E(θ, ϕ) radiated by two dipoles (h, ±d, 0) located above a rectangular screen for each components of electric vector Eθ,ϕ (θ, ϕ) is defined in common coordinate frame XYZ as the sum of two independent fields E 1 and E 2 determined in frame X 1 Y 1 Z 1 for dipole 1 and in frame X 2 Y 2 Z 2 for dipole 2 by the following formula ⎛ ⎞ 4   E = ejkD1 ⎝ Ei χi + Ein χin + E12 χ2 + E21 χ21 + E34 χ34 + E43 χ43 ⎠ i=1,3 n=1

i=1,3

⎛

+ e−jkD1 ⎝



Ei χi +

i=2,4

4 

⎞     Ein χn + E12 χ12 + E21 χ21 + E34 χ34 + E43 χ43 ⎠.

i=2,4 n=1

(1) In (1) k = 2π/λ is a wave-number in free space, D1 is the path-length difference of the rays going to the observation point from points O and O1 (Fig. 1); and D1 = d sinθsinϕ. The first terms in the parentheses determine the sum of geometric optics (GO) fields E i , where i = 1, 3 are the numbers of the sources of the incident wave from dipole 1 and its mirror image and i = 2, 4 are the numbers of the sources of the incident wave from dipole 2 and its mirror image; the second terms in the parenthesis determine the sum of first-order diffracted fields Ein that are excited by the GO fields from the four sources with numbers i = 1 ÷ 4 on the four edges of the screen n = 1 ÷ 4; and the  , E  , E  , E  are determined by the waves diffracted fields E12 , E21 , E34 , E43 and E12 21 34 43 twice by the parallel edges of the screen in case of dipole 1 and 2 correspondingly. Since the screen sizes are finite, each field has its own regions of light and shadow. In (1) the coefficients χi and χn are equal to one and zero in the light and shadow regions of each field, whose boundaries are determined in frame R, θ, ϕ. The components of diffracted field are defined in the rectangular proper coordinate systems Xn Yn Zn on the n-th screen edge (Fig. 1), and then are projected on the axis of frame R, θ, ϕ. All numerical results are presented in the frame R , θ , ϕ with polar axis Z normal to the screen (Fig. 1), in ˙  (θ , ϕ ), E ˙  (θ , ϕ ) can be written as which the components of electric field vector E θ ϕ  θ (θ , ϕ ) = [Eθ (θ, ϕ)C1 + Eϕ (θ, ϕ)C2 ]θ , E  ϕ (θ , ϕ ) = [−Eθ (θ, ϕ)C2 + Eϕ (θ, ϕ)C1 ] E ϕ , 





C1 = −cos θ cos ϕ/ sin θ , C2 = sin ϕ/ sin θ , sin θ =



1 − sin2 θ cos2 ϕ.

(2)

The Radiation Characteristics of Two Coupled Vertical Dipoles

345

3 Determination of Radiation Patterns 3.1 Electrical GO Field Strength Vector GO Field of the Dipole 1 Placed Perpendicularly to the Infinite Plane, in the Coordinate System X 1 Y 1 Z1 (R, θ, ϕ). When the infinite plane is excited by an symmetric dipole with the arm length l oriented along the X 1 axis (Fig. 1), the GO field strength vector of the incident (i = 1) and reflected from the plane (i = 2) waves in a spherical coordinate system R, θ, ϕ, associated with the system X 1 Y 1 Z 1 , in the far zone has two components

Eθ i (θ, ϕ) = − E0 F⊥ (θ, ϕ) cos ϕ cos θ exp (jkδi ), Eϕi (θ, ϕ) = E0 F⊥ (θ, ϕ) sin ϕ exp(jkδi ), cos(kl sin θ cos ϕ) − cos kl , F⊥ (θ, ϕ) = 1 − sin2 θ cos2 ϕ

(3a)

, I 0 is a current value in the dipole middle, δi is the path-length where E0 = j60I0 exp(−jkR) R difference of the rays going from the origin O1 of the frame X 1 Y 1 Z 1 and from the phase center of the i-th emitter with polar coordinates (r0i , ϕ0i ) to the observation point M (θ, ϕ), δi = r0i sin θ cos(ϕ − ϕ0i ), ϕ01 = 0; ϕ02 = π; r0i = h. In case of a half-wave dipole, we have  π sin θ cos ϕ / 1 − sin2 θ cos2 ϕ . (3b) F⊥ (θ, ϕ) = cos 2 After summing the fields (3a) and normalizing to the E 0 , with account formula Euler cos z = (ejz + e−jz )/2,

(4)

the patterns of the components of the total electric field of the dipole located perpendicularly to the infinite screen at a height h above it in the frame X 1 Y 1 Z 1 are fϕ = 2F⊥ϕ cos(kh sin θ cos ϕ),

(5a)

fθ = −2F⊥θ cos(kh sin θ cos ϕ),

(5b)

F⊥θ (θ, ϕ) = F⊥ (θ, ϕ)cos ϕ cos θ, F⊥ϕ (θ,ϕ) = F⊥ (θ,ϕ)sin ϕ,

(5c)

where

GO Field of Two Coupled Dipoles Placed Perpendicularly to the Infinite Plane, in the Frame XYZ (R, θ, ϕ). As a model solution of the problem obtain analytical expression for the radiation pattern of two identical dipoles 1 and 2 with arm length l and the same current I0 located at a distance ξ = 2d from each other at a height h above an ideally conducting infinitely thin plane, in the frame R, θ, ϕ.

346

N. Yeliseyeva et al.

The electrodynamics model of the GO field in accordance with expression (1) is an array of two pairs of parallel emitters with co-phase currents of the same amplitude, therefore the amplitude of the total field of dipole 1 and 2 is equal to the sum of the amplitudes of the GO fields of all four emitters (i = 1, 2, 3, 4) Eθ(1,2) (θ, ϕ) = −E0 F⊥θ (θ, ϕ)

4 

exp(jkδi ), Eϕ(1,2) (θ, ϕ) = E0 F⊥ϕ (θ, ϕ)

i=1

4 

exp(jkδi ),

i=1

(6) where Fθ⊥ (θ, ϕ), Fϕ⊥ (θ, ϕ) are defined by formulas (5c), δi is the path-length difference of the rays going from the origin O of the coordinate system XYZ and from the phase  , ϕ ) to the observation point M (θ, center of the i-th emitter with polar coordinates (r0i 0i ϕ). Here i = 1, 2 are the numbers of the sources of the incident waves from the dipoles 1 and 2, i = 3, 4 – numbers of their mirror images respectively. The polar coordinates of the sources i = 1, 2, 3, 4 in the coordinate system R, θ, ϕ (Fig. 1) are equal

 (7) ϕ 01,2 = arctg ±d h ; ϕ 03,4 = π ∓ ϕ 01 ; r0 = h2 + d 2 , where the upper signs refer to sources i = 1, 2, the lower signs for sources i = 3, 4. The value in front of the sum sign in (6) is the field of the emitter at the origin O of the XYZ coordinate system, and the sum is an interference factor of the co-phase emitters system. The path-length difference of the rays going from the origin O of the   frame XYZ and from the phase center of the i-th emitter

with polar coordinates (r0i , ϕ0i )   to the observation point M (θ, ϕ) δi = r0i sin θ cos ϕ − ϕ0i are equal



  δ1 = r0 sin θ cos ϕ − ϕ01 ; δ2 = r0 sin θ cos ϕ + ϕ01 ;



      δ3 = −r0 sin θ cos ϕ + ϕ01 ; δ4 = −r0 sin θ cos ϕ − ϕ01 . (8) After substituting (8) in (6), normalizing the fields on the E 0 , using the Euler formula  and d = r  sin ϕ , we obtain the radiation (4), and with account that h = r0 cos ϕ01 0 01 patterns for the components of the total field from four co-phase emitters in the form fθ(1,2) (θ, ϕ) = −4F⊥θ (θ, ϕ) cos(kh sin θ cos ϕ) cos(kd sin θ sin ϕ),

(9)

fϕ(1,2) (θ, ϕ) = 4F⊥ϕ (θ, ϕ) cos(kh sin θ cos ϕ) cos(kd sin θ sin ϕ). From a comparison of (9) and (6) it follows that the patterns of a system of two vertical coupled dipoles with infinite plane in the frame R, θ, ϕ are equal to the product of the patterns of a single dipole F⊥θ,ϕ (θ, ϕ) located in the origin O of the frame R, θ, ϕ by the interference array factor of four co-phase emitters with δi (8) N(1,2) (θ, ϕ) = 4 cos(kh sin θ cos ϕ) cos(kd sin θ sin ϕ).

(10)

GO Field of the Dipole 1 Placed Perpendicularly to the Infinite Plane, in the Frame X Y Z (R , θ , ϕ ). In this frame a vertical dipole has one field component

The Radiation Characteristics of Two Coupled Vertical Dipoles

347

Eθ i θ = E0 Fθ (θ ) exp(jkδi ),

(11a)



 Fθ θ = cos kl cos θ − cos kl / sin θ ,

(11b)

where δi is the path-length difference of the rays going from the origin O of the frame X Y Z and from the dipole 1 (i = 1) and its mirror image (i = 2) to the M (θ , ϕ ).   (θ , ϕ ) from O to M (θ , ϕ ) has the coordinates In this frame vector R X  = R cos θ cos ϕ , Y  = R cos θ sin ϕ , Z  = R cos θ .

(12)

The coordinates of the dipole and its mirror image are x = 0, y = 0, zi = ±h,   (θ , ϕ ) so the path-length differences δi of the rays coming from them to M (θ , ϕ ) and R equal to δi = ±h cos θ .

(13)

The radiation pattern of total field of the incident and reflected waves from the infinite plane with account for (11a, 11b)–(12) and (4) in the frame R , θ , ϕ has form fθ (θ , ϕ ) = 2Fθ (θ ) cos(kh cos θ ).

(14)

GO Field of Two Coupled Dipoles Placed Perpendicularly to the Infinite Plane, in the Frame X Y Z (R , θ , ϕ ). Now obtain the analytical expression for the radiation pattern of two identical dipoles 1 and 2 located at a distance ξ = 2d from each other at a height h above a plane, in frame R , θ , ϕ . In this case the field is excited by an array of four co-phase dipoles (i = 1, 2, 3, 4). The field of each from them is

Eθ i θ , ϕ = E0 Fθ (θ ) exp(jkδi ), (15) where δi is the path-length difference of the rays going to the observation point M (θ , ϕ ) from point O and from i-th dipole with coordinates (0, ±d , ±h) (Fig. 1). We have δi = (X  x  i + Y yi + Z z i )/R , and with account of (12) in following forms δ1 = d sin θ sin ϕ + h cos θ , δ3 = d sin θ sin ϕ − h cos θ ,

(16)

δ2 = −d sin θ sin ϕ + h cos θ , δ4 = −d sin θ sin ϕ − h cos θ , The radiation pattern of the total field of four dipoles normalized to the E 0 fθ (1,2) (θ , ϕ )





= F (θ , ϕ ) θ

4 

exp(jkδi )

i=1

we obtain after substituting (16) in (15) and using the Euler formula (4) in the form







(17) fθ (1,2) θ , ϕ = 4Fθ θ , ϕ cos kh cos θ cos kd sin θ sin ϕ .

348

N. Yeliseyeva et al.

From a comparison of (11b) and (17) it follows that the pattern of two vertical    of the coupled dipoles with infinite plane

 in the frame R , θ , ϕ is equal to the product  pattern of a single dipole Fθ θ , ϕ located in the origin O of the frame R, θ , ϕ by the interference array factor of four co-phase emitters with δi (16)

 



 N(1,2) θ , ϕ = 4 cos kh cos θ cos kd sin θ sin ϕ .

(18)

The pattern of two coupled half-wave dipoles 1 and 2 over the infinite plane takes form

  

π

   cos kh cos θ cos kd sin θ sin ϕ  fθ (1,2) θ , ϕ = 4 cos cos θ . (19) 2 sin θ So from (19) in the main observation planes we have cos kh cos θ cos kd sin θ π

  ◦  fθ (1,2) θ , ϕ = 90 = 4 cos cos θ 2 sin θ 

π

  ◦  cos kh cos θ  fθ (1,2) θ , ϕ = 0 = 4 cos cos θ . 2 sin θ

(20a) (20b)

Thus, the radiation patterns in the main planes (20a, 20b), in contrast to the case with one dipole above the plane (14), are not symmetrical. It can be seen from (20a) and (20b) that it is possible to ensure the same width of the main lobe of the RPs in the main observation planes by controlling the shape of the RP of the E θ -component in the plane ϕ = 90°, 270° due to changes in the distance between the dipoles ξ = 2d. It’s seen from (20b) that in the observation plane ϕ = 0, the radiation pattern as well as in the case of two parallel dipoles to the infinite screen [3], doesn’t depend on the distance ξ = 2d between the dipoles. The maximum amplitude reaches 4E 0 in the plane of the screen when θ = 90° regardless of the distance h between the screen and the dipole. Amplitude of E θ -component is equal to twice the amplitude of the field of a one vertical dipole above the plane. So, the normalized patterns in planes ϕ = 0°, 180° for one dipole placed above screen midpoint and two dipoles coincide. The analysis of the patterns for vertical dipole with a rectangular screen depending on the screen sides’ ratio W/L, size L/λ, dipole removal h/λ from the screen was carried out in [4]. With account that the radiation patterns in the planes ϕ = 90°, 270° do not depend on the screen shape W/L, we’ll analyze the patterns in the main planes in dependence of the distance between the dipoles ξ for the square screens with different L/λ. 3.2 Diffracted Field Strength Vector The far-zone diffracted fields that are excited by the radiation of dipole 1 on the screen edges are determined by the patterns of virtual diffraction radiators (VDR) placed at the origins On of the proper coordinate system (PCS) x 1 y1 z1 , x 2 y2 z2 , x 3 y3 z3 and x 4 y4 z4 on screen edges. The diffracted fields from dipole 2 are determined by the VDRs, placed at the origins of PCS x 1 y1 z1 , x 2 y2 z2 , x 5 y5 z5 and x 6 y6 z6 (Fig. 1). When the screen edges are excited by the dipole i = 1, 2, 3, 4 oriented perpendicular to them, the primary diffraction

The Radiation Characteristics of Two Coupled Vertical Dipoles

349

field strength vector in PCS Rn , θn , ϕn on the edge n = 1, 2, 3, 4 has two components described by the uniform in the observation angles asymptotics √ 

0n 2 sin ϕn + ϕn ∓ϕ 2 Eθin (θn , ϕn ) = Eθi T (ξin ) ∓ √ πkr0n sin θn  

 × cos θn exp −j kr0n sin θn + π 4 − kδn , (21a) √



0n 2 cos ϕn ± ϕn ∓ϕ 2 Eϕin (θn , ϕn ) = Eϕi T (ξin ) − √ πkr0n sin θn  

 × exp −j kr0n sin θn + π 4 − kδn .

(21b)

Here Eθi (θn , ϕn ) and Eϕi (θn , ϕn ) are the GO field components defined in the system Rn , θn , ϕn using formulas (3a)–(3b); the upper signs are taken at i = 1, 2 and lower signs at i = 3, 4; δn is the path-length difference of the rays passing from the origin O and from the origin On of the frame Rn , θn , ϕn in the coordinate system R, θ, ϕ δn = x0n sin θ cos ϕ + y0n sin θ sin ϕ + z0n cos θ,

(21c)

where x0n , y0n , z0n are the rectangular coordinates of the PCS origin On in the XYZ. The uniform field asymptotics for the edge waves (21a, 21b) contain the transition function from light to shadow T (ξin ) for the i-th wave diffracted at n-th screen edge

√ ξin  

 |ξin | j − 1 −1/2 exp(jπ/4) exp −jt 2 dt (22) T (ξin ) = ± ,  |ξin | j = 2π 2 0

where “+” and “−” signs are used respectively in the light and shadow regions of the GO field, the argument of probability integral  equal

ξin = 2krin sin θn cos((ϕn − ϕin )/2), rin , ϕin are the polar coordinates of i-th dipole in the frame Rn , θn , ϕn . The secondary diffraction half-shadow fields on the parallel edges of the screen AD and BC (m = 2, n = 1), BC and AD (m = 1, n = 2), AB and CD (m = 3, n = 4), CD and AB (m = 4, n = 3) (Fig. 1) are described by the uniform at observation angles asymptotic  

d ± |ξmn | j − 1 (23) Eθ,ϕ(mn) = ±Eθ,ϕm ξ12 =

√

√



2kL sin θ sin 45◦ − ϕ/2 , ξ21 = 2kL sin θ cos 45◦ − ϕ/2 ,

ξ34 =

2kW cos ϕ cos(θ/2), ξ43 =

2kW cos ϕ sin(θ/2),

where the first “+” sign is taken in the light region of the GO field, the second in the light region of the primary diffraction field; the sign “–” is taken in the shadow areas d – uniform asymptotic of the components of the primary of one or another field; Eθ,ϕm diffraction field from the edge m, determined by the expressions (21a)–(21b).

350

N. Yeliseyeva et al.

4 Analysis of the Computation Results 4.1 Analysis of the Calculated Patterns On the base of the asymptotic solution (1) the fast 3D algorithms and software have been developed for computation of the radiation patterns (RPs) for two vertical dipoles placed over the rectangular ideal conducting screen as function of the geometric parameters and electric dimensions of the radiating system in the whole space. The all computation results are presented in case of half-wave dipoles. Figure 2 illustrates the RPs of the field at the distance between dipoles 1 and 2 ξ = 0.5λ and dipole’s removal from the screen h = 0.3λ in case of square screen with L = W =λ. In Fig. 2a normalized to E0 pattern of E θ -full field component (|Eθ | - curve  1) Eθgo  is presented in the planes ϕ = 0°, 180° in the parts: the patterns of GO field (  curve2) and of diffracted fields on edges 1–4 (Eθdif  - curve 3). The boundary angles of the light-shadow regions for the incident and reflected GO waves concerning direction θ = 90° (a screen plane) in plane ϕ = 0° equal to ±30°, i.e. GO field exists at angle sector θ = 0° ÷ 120°, and is a superposition of incident and reflected waves at θ = 0° ÷ 60°. In Fig. 2b the patterns of full field Eθ (θ , ϕ ) normalized to maximum in the whole space amplitude E θmax = 2.23 E0 are presented in plane ϕ = 0° (solid  curve   = 90°(solid curve 2). The spatial distributions of diffracted fields E  1) and ϕ θdif (c),     Eϕ dif (e), GO field Eθgo  (d) and full field |Eθ | (f) are shown in the frame of angles ϕ , θ . In Figs. 2g, h the patterns of the Eθ -component at the ξ = 0.5λ and h = 0.3λ in case of square screen with L = W = 3λ are presented. It is seen that RPs of Eθ -component of total field at ξ = 0.5λ and h = 0.3λ has two symmetric peaks-radiation (Fig. 2f, h) in the plane ϕmax = 0° at θmax = 74° and θmax = 67° in case of the screen with L = λ and L = 3λ accordingly. Let’s pass to consider the formation of diffracted field (Fig. 2c, e) in the main observation planes and in the whole space. During diffraction on the finite size screen, in the observation space the shadow cones of diffracted waves with vertices at the ends of the edges are formed [6]. In Fig. 1 the shadow cones βn of diffracted waves from screen edges n = 1, 2 (β1, 2 ) and n = 3, 4 (β3, 4 ) are shown. The aperture angles of shadow cones for dipoles 1 and 2 placed over rectangular screen are defined as ⎛  ⎛ ⎞ ⎞ 2 h2 + L21,2 h2 + (W /2)2 ⎠, ⎠, β3,4 = arctg⎝ β1,2 = arctg⎝ (24) W L1,2 where L1 = L/2 − d , L2 = L/2 + d for dipole 1 and L1 = L/2 + d , L2 = L/2 − d for dipole 2. Note that the region of shadow of diffracted fields from edges n = 3, 4 (χ3, 4d = 0) in the observation plane ϕ = 90° includes angles θ 90◦ − β3 < θ < 90◦ + β3 ,

(25)

The region of shadow of diffracted fields from edges n = 1, 2 (χ1, 2d = 0) in the observation planes ϕ = 0° is formed at angles θ 90◦ − β1 < θ < 90◦ + β1 .

(26)

The Radiation Characteristics of Two Coupled Vertical Dipoles

351

Fig. 2. RPs of the field components in the main planes and the spatial field distribution in the frame ϕ , θ at ξ = 0.5λ and h = 0.3λ in cases of square screens with L = λ and L = 3λ.

As a result of crossing the shadow cones with observation sphere in the far zone the circles are formed. In Fig. 2c, e one can see projections of the shadow cones from the

352

N. Yeliseyeva et al.

edges 1, 2 in the form of circles of radius β1, 2 with the centre at angles θ = 90°, ϕ = 0, 180°, 360°, and from the transverse edges 3, 4 in the form of circles of radius β3, 4 with the centre at θ = 90°, ϕ = 90, 270°. Cross-polarized Eϕ (θ , ϕ ) component is defined only by diffracted field (Fig. 2e); it is equal to zero in the main observation planes. Figure 3 illustrates the influence of the phase shift  between dipoles currents on their patterns. In case of the screen with L = W = 3λ at ξ = h = 0.25λ the patterns of the Eθ -component are shown for in-phase dipoles  = 0° (a, b) and for dipoles with  = 90° (c, d) in the planes ϕ = 0° and ϕ = 90° (a, c) and the spatial field distribution |Eθ | in the frame ϕ , θ (b, d). The in-phase dipoles are formed almost omnidirectional radiation with four peaks-radiation (b) at θmax = 65° symmetrically of ϕ = 0° (E θmax = 3.19 E0 ). In case of  = 90° (d) the radiation is unidirectional (symmetrically of the plane ϕ = 90°) with the peak at θmax = 61°, ϕmax = 50° and E θmax = 3.59 E0 . Figure 3 shows also the patterns of the Eθ -component in case of the same screen L = W = 3λ and h = 0.25λ but at ξ = λ in the planes ϕ = 0° and ϕ = 90° (e) and the spatial field distribution |Eθ |(d). In this case are formed four peak radiation with E θmax = 3.18 E0 at θmax = 64° symmetrically of the planes ϕ = 0, 90°, 180, 270°. For verification of the developed algorithm and the numerical results in the Figs. 2 and 3 here are also presented the patterns (dotted curves) of the same radiating systems calculated using commercial software Feko. The 3D patterns of Eθ ,ϕ -components calculated by software Feko for analyzed radiating systems are presented in Fig. 4. The patterns in the main planes and the peak angles of the 3D radiation are in a good agreement with numerical results obtained by method of UGTD beginning from the screen with L = W = λ at h = 0.3λ. 4.2 Analysis of the Directive Gain and the Radiation Resistance Next we’ll analyze the directive gain D(θmax , ϕmax ) in the direction of maximum radiation and the radiation resistance R of two half-wave vertical dipoles 1 and 2 placed above the perfectly conducting screen (Fig. 1) depending on the screen width L/λ and the screen sides ratio W/L at different distances between dipoles ξ/ λ and constant dipoles removal from the screen h/ λ. The radiation resistance R and directive gain D(θ , ϕ ) are defined through the mean radiated power of an antenna P R2 P = 240π

2π π  dϕ 0

 2  E  sin θ dθ .

(27)

0

2  2  Substituting |E|2 in (27) in the form |E|2 = E02 f (θ , ϕ ) , where f (θ , ϕ ) is the normalized power pattern for two dipoles with the screen 2  2   |f |2 = fθ1 (θ , ϕ ) + fθ2 (θ , ϕ ) + fϕ1 (θ , ϕ ) + fϕ2 (θ , ϕ ) , (28) fθ,ϕ1 (θ , ϕ ), fθ,ϕ2 (θ , ϕ ) are the radiation patterns of the orthogonal linearly polarized field components in case of exciting a screen by dipole 1 and dipole 2, we obtain P = AI |I0 |2 /2.

(29)

The Radiation Characteristics of Two Coupled Vertical Dipoles

353

Fig. 3. RPs of E θ components in the main planes and the spatial field distribution in the frame ϕ , θ at L = 3λ: at ξ = h = 0.25λ for  = 0° (a, b) and  = 90° (c, d); at ξ = λ and h = 0.25λ (e, f)

Here I 0 is a current value in the dipole middle, A is numerical factor and the radiation integral is 2π π    2  f (θ , ϕ ) sin θ d θ I = d ϕ 0

0

(30)

354

N. Yeliseyeva et al.

Fig. 4. The 3D patterns of E θ (a, b, c, d) and E ϕ (e, f) components calculated by software Feko: (a) L = 3λ, ξ = λ, h = 0.25λ, (b) L = 3λ, ξ = h = 0.25λ, (c, e) L = 3λ, ξ = 0.5λ, h = 0.3λ, (d, f) L = λ, ξ = 0.5λ, h = 0.3λ.

Based on the definition (29) we obtain the radiation resistance in the form R = AI , with account that for symmetrical dipole of length 2l numerical factor is A = 30/π, R = 30I /π . The directive gain of the antenna is defined as



D θ , ϕ = 4πP1 θ , ϕ /P ,

(31)

(32)

The Radiation Characteristics of Two Coupled Vertical Dipoles

355

where P (θ , ϕ ) and P1 (θ , ϕ ) are the total radiation power and the radiation power, referred to a unit solid angle in the far zone in a certain direction. From the definition (32) the directive gain takes the form  )/I , D(θ , ϕ ) = 4πf 2 (θ , ϕ )/I , Dmax = 4πf 2 (θmax , ϕmax

(33)

where θmax , ϕmax are the observation angles in the direction of the main radiation maximum. On base of calculation of the 3D radiation patterns fθ,ϕ (θ , ϕ ) of the orthogonal field components and using the expressions (31) and (33), the values of the directive gain Dmax and radiation resistance R were calculated at h = 0.3λ and the distances between dipoles ξ/ λ = 0.25, 0.5 and 1, when changing the screen width L/ λ = 1 ÷ 4 and the sides ratio W/L = 0.2 ÷ 2 with the step 0.05. The maximum Dopt and corresponding Ropt and the optimal screen side ratio (W/L)opt versus L/ λ have been determined. In Figs. 5, 6 and 7 the numerical results are presented: the equal lines of directive gain D(θmax, ϕmax ) at the maximum radiation (a) and radiation resistance R (b) in the frame W/L and L/λ; D(θmax , ϕmax ) (c) and R (d) vs. the screen width L/λ at W/L = 0.5, 0.75, 1, 1.5, 2 (curves 1–5) for ξ/λ = 0.25, 0.5 and 1. In Figs (c, d) the values of D(θmax, ϕmax ) and R in case of two vertical dipoles placed above infinite plane calculated by

Fig. 5. The equal lines of values D(θmax, ϕmax ) (a) and R (b) in the frame W/L and L/λ; D(θmax, ϕmax ) (c) and R (d) vs. L/λ at W/L = 0.5, 0.75, 1, 1.5, 2 (1–5) for ξ = 0.25λ, h = 0.3λ

356

N. Yeliseyeva et al.

formulas (19), (31) and (33) are shown too (line 6). In Fig. 6c the D(θmax , ϕmax )-values at W/L = 2 calculated by soft ware Feko are shown by points located near the curve 5.

Fig. 6. The equal lines of values D(θmax, ϕmax ) (a) and R (b) in the frameW/L and L/λ; D(θmax, ϕmax ) (c) and R (d) vs. L/λ at W/L = 0.5, 0.75, 1, 1.5, 2 (1–5) for ξ = 0.5λ, h = 0.3λ

Figures 8 illustrate the maximum values of the directive gain Dopt (a) at the observation angles θopt (c), and corresponding optimal radiation resistances Ropt (b) and the screen sides ratios (W/L)opt (e) depending on L/λ at ξ/λ = 0.25, 0.5, and 1 (curves 1, 2, 3). The optimal values are obtained from Figs. 5, 6 and 7(a, b). It can be seen that the directive gain sufficiently depends on the distance between the dipoles ξ and the screen sides ratios. So, the appropriately chosen distance ξ and the appropriately chosen (W/L)opt provide for the given patterns with high directive gain. Dopt reaches 12.2 at θopt = 70°, ϕopt = 11° when L = 3λ, (W/L)opt = 2, Ropt = 134 Ohm.

The Radiation Characteristics of Two Coupled Vertical Dipoles

357

Fig. 7. The equal lines of values D(θmax, ϕmax ) (a) and R (b) in the frame W/L and L/λ; D(θmax, ϕmax ) (c) and R (d) vs. L/λ at W/L = 0.5, 0.75, 1, 1.5, 2 (1–5) for ξ = λ, h = 0.3λ

Fig. 8. The maximum Dopt -values (a) at the angles θopt (c) and corresponding optimal Ropt (b) and the sides ratios (W/L)opt (d) vs. L/λ at ξ/λ = 0.25, 0.5, and 1 (curves 1, 2, 3).

358

N. Yeliseyeva et al.

5 Conclusions The 3D vector problem of diffraction of the fields of two coupled parallel electric dipoles placed in perpendicularly over an infinitely thin rectangular ideally conducting screen is considered. On the base of the asymptotic solution to this problem by the method of the uniform geometric diffraction theory, fast algorithms and software are developed for computation of the patterns, directive gain and radiation resistance as function of the geometric parameters and electric dimensions of the radiating system. It is shown that, the radiation patterns in the main planes, unlike the case with a single vibrator above the square screen are not symmetrical. When the distance between dipoles and screen is fixed, the appropriately chosen distance between the dipoles and the appropriately chosen dimensions of the screen provide for radiation with high directive gain. When the dipoles currents are in quadrature the patterns may be unidirectional. For verification of the calculations, the patterns and directive gains of the same radiating systems were calculated using commercial software Feko. The patterns in main observation planes are in a good agreement. The developed algorithm and software allow simulating effective real wireless communication systems and are thus useful for the further development of wireless communication systems and for the formation of new standards. The results are important for specialists in field of wireless communications.

References 1. Balanis, C.A.: Antenna Theory: Analysis and Design. 4th edn. Wiley, New York (2016) 2. Gorobets, N.N., Yeliseyeva, N.P., Antonenko, Ye.A.: Optimisation of radiation characteristics of wire-screened antennas. J. Telecommun. Radio Eng. 71(1), 59–69 (2012) 3. Gorobets, N.N., Yeliseyeva, N.P.: The radiation characteristics of two parallel electric dipoles with a finite size screen. J. Commun. Technol. Electron. 58(8), 753–761 (2013) 4. Yeliseyeva, N.P., Berdnik, S.L., Katrich, V.A.: Influence of sizes and sides ratio of rectangular screen on directive patterns of vertical dipole. In: 39th IEEE International Conference on Electronics and Nanotechnology (ELNANO 2019), pp. 745–749 (2019) 5. Yeliseyeva, N.P., Berdnik, S.L., Katrich, V.A.: Radiation patterns of two coupled vertical electric dipoles placed over rectangular screen. In: Proceedings of the 4th UkrMiCo 2019, Odessa, Ukraine, September 9–13, 2019 6. Keller, J.B.: Geometrical theory of diffraction. J. Opt. Soc. Amer. 52, 116–130 (1962)

The Analysis of Distributed Two-Layers Components in Three-Layer Planar Structure Yulia Rassokhina(B)

, Vladimir Krizhanovski(B)

, and Vasyl Komarov(B)

Vasyl’ Stus Donetsk National University, Vinnytsia, Ukraine [email protected], [email protected], [email protected]

Abstract. Distributed two-layer components (also known as defective ground structures) are widely used to create filters with a wide stop-band, including matching networks for high-efficiency amplifiers. They are usually calculated either by equivalent circuit methods or by numerical methods. For a better understanding of physical processes in a distributed two-layer discontinuity in three-layer structures, it is desirable to develop semi-analytical methods that allow us to obtain dependences of constant propagation on parameters of structure. The transverse resonance technique (TRT) is close to such methods, which calculates the scattering characteristics of discontinuity in planar-type transmission lines in the microwave range. It is based on the solution of boundary value problems for three-dimensional resonators containing, in the general case, multiplane discontinuity. Such name of this method is due to introduction of boundary conditions in the transverse with respect to the discontinuity of the direction in terms of the coefficients of reflection from the ideal boundaries (electric or magnetic). The key to effectively solving a boundary-value problem (for example, by the Galerkin method) is its algebraization method that is, the choosing a basis on which the unknown field or current components are decomposed. Keywords: Microstrip line · Step discontinuity · Complex-shaped slot resonator · Eigen frequency spectrum · Scattering characteristics

1 Introduction The spectral approach for solution of boundary problems in micro- and millimeter wave range devices design was developed since 80-th years of last century [1]. The analysis of multilayer planar structures in spectral domain allows define boundary problem in the form of integral equation, which comes down to solution of lower order linear equation system, and provides calculation of structural components in passive devices being designed. In [2], the resonance frequencies technique calculation for multilayer planar structure consisted of microstrip line resonator at coupled lines and tunable slot resonator in second plane is described. This technique is based on field description of shielded structure in spectral domain in the form of hybrid waves and boundary problem solution by Galerkin’s method. The Green’s functions formulated for the area being analyzed are © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 359–376, 2021. https://doi.org/10.1007/978-3-030-58359-0_20

360

Y. Rassokhina et al.

universal and applied to various microstripstrip-slot structures. The generalized analysis and design technique for multilayer components based on combination of spectral approach (using immittance matrix) and standard CAD methods is described in [3–6]. The interest of researchers in multi-layer planar structures, including DGS-structures (defected ground structure), does not weaken, on their basis, differential bandpass filters, directional coupler and vertical transitions are designed [7–9], microwave bandpass filter in a triplet configuration with a frequency-dependent cross-coupling [10], two types of trisection bandpass filters (BPF) with a mixed cross-coupling [11], symmetric trisection bandpass filter (BPF) with reflection-type half-wave resonators [12]. The transverse resonance method (TRT), by which the scattering characteristics of discontinuities in planar-type transmission lines are calculated in the microwave range, is based on the solution of boundary value problems for three-dimensional resonators containing such (in the general case, multilayer) discontinuity [3, 13, 14]. The denomination of the method is caused by the application of boundary conditions in the transverse (with respect to the discontinuity) direction in terms of the coefficients of reflection from the ideal electrodynamic boundaries (electrical or magnetic). The key moment in the solution of the boundary value problem by the Galerkin’s method is the method of its algebraization. A description of the current density in a regular microstrip transmission line by power functions (polynomials) was proposed in [15, 16], and a description of the current density and the field at the slot by such functions that take into account the behavior of the field on an infinitely thin edge for both microstrip and slot transmission lines, proposed in [14]. The use of Chebyshev polynomials of the first Tn (x) and second Un (x) kind, which have weight functions that correspond to the field behavior on a thin edge, was proposed further [17] to describe currents and fields in planar transmission lines. In an irregular transmission line, such as inductive or capacitive segments in a strip transmission line, the description of current density is complicated. This is due to the fact that orthogonal polynomials and trigonometric functions that describe an electromagnetic field in a shielded area with rectangular boundaries correspond to different differential equations. In [16] we used expansion in orthonormal basis of trigonometric functions, which satisfies the Helmholtz equation in [16] to construct the current density function in an irregular microstrip resonator. This algorithm converges more slowly (i.e., has a higher order of the system of linear algebraic equations) than algorithms in which the current density is described via sequence of orthogonal polynomials, which is its disadvantage. The algebraization of boundary problems for volume resonator with slot resonators discontinuities (narrow rectangular or complex shaped slot resonators) in microstrip line ground plane is carried out using the waveguide basis functions Th(e)y,k , k = 1..Ns , by which the field at aperture of slot resonator is decomposed. In [19] considered in the periodic structure in the narrow rectangular slot resonators in a microstrip line ground plane. In the paper [20], the scattering characteristics of fundamental wave in microstrip line at H-shaped slot resonators with various orientation relative to microstrip line were considered. It was shown, that the such discontinuities provide two-frequency, particularly, wideband attenuation, caused by the discontinues interaction at relatively big

The Analysis of Distributed Two-Layers Components

361

distances from each other. The scattering characteristics of multilayer distributed discontinuities included step discontinuity in microstrip line (inductive or capacitive sections) and narrow rectangular slot resonators were considered in papers [21]. The aim of this work is improving the design technique and investigation of the scattering characteristics resonance properties at a two-layer distributed discontinuity in three-layer planar structure consisted the step discontinuity in microstrip line and complex shaped slot resonators.

2 Transverse Resonance Method for a Three-Layer Planar Structure with a Two-Layer Discontinuity Transverse resonance technique is based on the solution of boundary problem for a resonator on a feed line that includes discontinuity and leads to the calculation of the scattering matrix elements [13]. Scattering matrix S of the 2-port network consisting of a microstrip line step discontinuity and complex shape slot resonator in its ground plane (Fig. 1) has the following form:   S11 S12 . (1) S= S21 S22

y

B b1

2

h

1 0

A

w/2 3

x -b2

as Fig. 1. Microstrip line with capacitive section and H-shaped slot resonator in its ground plane: symmetric three-layer structure.

For analysis of the structure which is symmetric with respect to z = 0 plane, for example a symmetric H-shaped slot resonator and capacitive section in microstrip line, Fig. 1, it is sufficient to solve two boundary problems: one under electric wall (e.w.) condition at the region’s boundary z = L = l1 + l2 and at the symmetry plane z = 0, and another with magnetic wall (m.w.) condition at the same planes. The characteristic system equation of the transverse resonance has the following form: 2 = 0, (S11 + 1 )2 − S12

362

Y. Rassokhina et al. 2 = 0, (S11 − 2 )2 − S12

(2)

  where 1(2) = exp 2j βz l2,1(2) , βz - propagation constant of the fundamental wave of the microstrip line. The elements of the scattering matrix S11 = S22 , S12 = S21 are calculated from the solutions of the “electric” (e.w.–e.w.) and “magnetic” (m.w.–m.w.) boundary problems with respect to resonator dimensions l2,k (k = 1..2 is the number of solution) according to the formulae obtained from solution of two equations: S11 =

2 − 1 1 + 2 , S12 = . 2 2

(3)

From (3) we conclude that the frequencies of resonant transmission (minimum of S 11 ) are determined by the intersection points of the spectral curves (when 1 = 2 ) from the solutions of two boundary problems. Then the elements of the scattering matrix S11 = S22 , S12 = S21 are calculated from the solutions of the “electric” (e.w.–e.w.) and “magneto-electric” (m.w.–e.w.) boundary problems with respect to resonator dimensions l2,k (k = 1..2 is the number of solution) 2 = 0, (S11 + 1 )2 − S12 2 = 0, (S11 + 2 )2 − S12

(4)

according to the formula obtained from solution of two Eqs. (4): S11 = −

2 + 1 1 − 2 , S12 = 2 2

(5)

  where 1(2) = exp 2j βz l2,1(2) . From (5) we conclude that the frequencies of resonant interaction are determined by the intersection points of the spectral curves (when 1 = 2 ) from the solutions of two boundary problems. Figure 1 shows an example of symmetric in the transverse direction (with respect to z = 0 plane) resonant structure in the form of H-shaped slot resonator in the microstrip line’s ground plane and capacitive section microstrip line (top view are shown and cross-section), i.e. two-layer distributing discontinuity in three-layer planar structure. The structure is symmetric with respect to x = 0 plane. The first layer (i = 1) represents a dielectric substrate with relative dielectric permeability εr1 and thickness h, the second and third layers are the air ones, εr2 = εr3 = 1. The field components satisfy the conditions of ideally conductive electric wall at x = ±A, y = B, y = −b2 and conditions of electric or magnetic wall at longitudinal boundaries z = ±L and symmetric plane z = 0 for “electric”, “magnetic” or “magneto-electric” boundary problems. The boundary problem for the resonator is solved by the partial regions method according to which the original region is split into 3 partial regions (Fig. 1), and for each partial region the Helmholtz equation is solved for electric e and magnetic h vector potentials 0, Ah(e)y , 0 : Ah(e)y,i + k02 εri Ah(e)y,i = 0

 where k0 = ω0 c is the wave number and εri is the relative dielectric permeability of the i-th layer.

The Analysis of Distributed Two-Layers Components

363

Electric and magnetic Hertz vector-potentials in the chosen coordinate system are represented as double Fourier series expansions of the following form      sin kzn z Fei,mn (y) = Pmn cos kxm x φmn (x, z)Fei,mn (y), Aey,i = cos kzn z m=1 n=1(0) m=1 n=1(0)      cos kzn z Pmn sin kxm x ψmn (x, z)Fhi,mn (y), Ahy,i = Fhi,mn (y) = sin kzn z m=1 n=0(1)

m=1 n=0(1)

(6) for “electric” and “magnetic” boundary problems, correspondingly, where i = 1…3 is the partial region number. The following notations were introduced here:



  2 2 2 Pmn = 2 A (2 − δn0 ) Lχ−1 mn , χmn = kxm + kzn , kxm = π(2m − 1) 2A,  where δn0 is the Kronecker symbol, moreover, kzn = πn  L for an “electric” or “magnetic” boundary value problem and kzn = π(2n − 1) 2L for the “magneto-electric” problem.  The functions Fe(h)i,mn (y) are written as follows (l = (m, n), y0 = h 2): Fe1,l = Re11l Fe2,l = Re2l

cos ky2l (B − y) cos ky2l (b2 + y) , Fe3,l = Re3l , ky2l sin ky2l b1 ky2l sin ky2l b2

Fh1,l = Rh11l Fh2,l = Rh2l

sin ky1l (y − y0 ) cos ky1l (y − y0 ) + Re12l , ky1l cos ky1l y0 ky1l sin ky1l y0

cos ky1l (y − y0 ) sin ky1l (y − y0 ) + Rh12l , cos ky1l y0 sin ky1l y0

sin ky2l (B − y) sin ky2l (b2 + y) , Fh3,l = Rh3l , sin ky2l b1 sin ky2l b2

2 2 − k 2 and R where kyi,mn = k02 εri − kxm e11(2),mn , Re2(3),mn , Rh11(2)mn , Rh2(3)mn are zn unknown coefficients of the Fourier series expansions. The boundary problem is solved by Galerkin’s method. In order to do so, the field in the slot resonator is written as series of eigenfunctions Th(e)y,k of the TE- and TH-waves of ridged waveguide:     E0t = Vhk ∇t Thy,k × ey + Vek ∇t Tey,k , k=1

H0t =

 k

Ihk ∇t Thy,k −



k=1

  Iek ∇t Tey,k × ey ,

(7)

k

where Vh(e)k are unknown coefficients of the series expansion. Current density in the strip line defined through the difference of tangential components of magnetic field in the y = h plane Ht,1 − Ht,2 = J × ey . By applying Galerkin’s procedure to continuity equations of the field’s tangential components on the partial regions’ boundaries y = 0 and y = h, obtain a homogeneous

364

Y. Rassokhina et al.

linear system of algebraic equations (SLAE) with unknown parameter (longitudinal dimension L = l1 +l2 of the resonator or its eigenfrequency fres ) with respect to unknown coefficients of the field expansion on the slot resonator and current density in the strip line. By equating the determinant of the linear system of equations to zero the characteristic equation for determining that parameter is obtained. The coupling integrals between the basic functions describing the electromagnetic field at the slot (7) and the basic functions of the three-dimensional resonator (6) are calculated by the formulas:  

  2 ∇ψmn (x, z) × ey ∇t Thy,k × ey dS = khc,k ψmn (x, z)Thy,k (x, z)dS, αhh,mn,k = S0

 αhe,mn,k = 

αeh,mn,k =

S0



∇ψmn (x, z) × ey ∇t Tey,k dS = 0,

S0

  ∇t φmn (x, z) ∇t Thy,k × ey dS =

S0

 dThy,k φmn dThy,k φmn − dS, dx dz dz dx

S0



αee,mn,k =

 

∇t φmn (x, z)∇t Tey,k dS =



χ2mn

S0

φmn (x, z)Tey,k dS.

(8)

S0

3 Solution of the Boundary Value Problem for Microstrip Resonators with Capacitive Discontinuity The topology of the strip resonator with capacitive (for certainty) discontinuity and transverse boundary conditions, for which the density distribution function of longitudinal and transversal currents is constructed, shown in Fig. 2. It also shows the decomposition of the output region of the strip resonator into two partial subregions. The conditions at the longitudinal boundaries, electric and magnetic walls (e.w. and m.w.), correspond to the resonators whose eigen frequencies are used to calculate the elements of the scattering matrix of the discontinuity by the transverse resonance method (TRT). In the following, boundary value problems with such boundary conditions will be called “electric” and “magnetic” boundary value problems, respectively. The notion of current density J flowing in the strip transmission line in the electrodynamics theory of the microwave range is introduced through the difference of tangent components of the magnetic field H on both sides of the microstrip transmission line: Ht,1 − Ht,2 = J × ey . Consider the solution of the boundary value problem for the current density J of a microstrip resonator, which is expressed by the vector potentials of the magnetic Jh and electrical Je types, namely, the longitudinal (z) and transversal (x) components of the current. They are calculated by the formula: Jt =

P 

∇Je,n (x, z) × ey Ce,n − n=1

P 1  ∇Jh,n (x, z)Ch,n , j · k0 n=1

(9)

The Analysis of Distributed Two-Layers Components

365

а

e.w. (m.w.)

-A

e.w. (m.w.)

-w1/2 -w2/2 2

1

0 w1/2

l1

m.w.

w2/2

l2

A L

x

b Fig. 2. (a) Topology of a strip transmission line for which boundary problems are solved with finite length step discontinuity w1 > w2 and (b) half of a symmetrical structure (strip line) with boundary conditions of electric (e.w.) and magnetic (m.w.) walls at longitudinal boundaries (transverse resonance conditions).

 where k0 = ω0 c, ω0 – circular frequency, c – speed of light in vacuum, Jh(e),n (x, z) – eigenfunctions of magnetic and electric vector potential for current density, Ch(e),n – unknown coefficients of series expansion, P – order of a series reducing (9). The components of a magnetic field in a shielded structure satisfy the Helmholtz equation, that is, the wave equation in Cartesian coordinates. But the function of current density in the microstrip transmission line has singularities at the edges of the strip (thin edge), so to describe the current density we will use first  Tn (x) and second kind Un (x) √ √ 1 − x2 and 1 − x2 that of Chebyshev polynomials that have weight functions 1 correspond to the singularities of the field behavior on the thin edge and satisfy their own differential equation. The condition of the magnetic wall in the plane x = 0 (Fig. 2) corresponds to the main wave of the strip (microstrip) transmission line in a symmetric structure, from which we obtain additional conditions for the functions of electric and magnetic vector potentials: Je,i (0, z) = 0,

dJh,i (0, z) = 0, dx

366

Y. Rassokhina et al.

where i = 1, 2 - number of partial domains. In view of this, the two-dimensional magnetic vector potential function Jh,n (x, z) in partial domains 1 and 2 is decomposed by Chebyshev orthogonal polynomials of the first kind of even order:     x x  T2k = cos 2k arccos  , k = 0, 1, 2 . . . wi 2 wi 2 satisfy the equation (i = 1, 2) ⎛ ⎝1 −



2 ⎞    2     2 d x 2k x x x ⎠ d T2k     T − + = 0, T   2 2k 2k wi 2 wi 2 wi 2 wi 2 wi 2 dx2 wi 2 dx x 

or in operator form (where L is a differential operator):     2   x x 2k    . T2k L T2k =− wi 2 wi 2 wi 2

(10)

(11)

The two-dimensional function of the electric vector potential Je,n (x, z) is decomposed in series by Chebyshev orthogonal polynomials of the second kind of odd order     u2k+1 wix/ 2 x  U2k+1 = 2 , k = 0, 1, 2 . . .  wi 2 x 1 − wi / 2     where u2k+1 wix/ 2 = sin (2k + 2) arccos wix/ 2 – second-order Chebyshev functions satisfying the same differential Eq. (10), or in the operator form:      2  x x 2n + 2    L u2n+1 u2n+1 =− wi 2 wi 2 wi 2 Then the differential equation with respect to the vector potentials of current density has the form:   d2 L Jh(e),n (x, z) + 2 Jh(e),n (x, z) + χ2h(e),n · Jh(e),n (x, z) = 0, dz

(12)

χ2h(e),n – eigenvalues corresponding to eigenfunctions Jh(e),n (x, z), n = 1, 2… In addition, the eigenfunctions satisfy the boundary conditions at the free boundaries of the inhomogeneous strip line:       dJhi ±wi 2, z = 0, Jei ±wi 2, z = 0, i = 1, 2. dx At the longitudinal boundaries z = 0 and z = L, the electric Je and magnetic Jh vector potentials of current density satisfy the following transverse boundary conditions: Jh1 (x, 0) = 0, Jh2 (x, L) = 0,

The Analysis of Distributed Two-Layers Components

367

d d Je1 (x, 0) = 0, Je2 (x, L) = 0 dz dz with perfectly conducting electric wall at z = 0 and z = L, and d Jh1 (x, 0) = 0, Jh2 (x, L) = 0, dz Je1 (x, 0) = 0, Je2 (x, L) = 0 with the ideal magnetic wall on them. In view of the above, the two-dimensional function for the magnetic vector potential in partial regions 1 and 2 is written as:     M   x 2 4 − 2 · δk0  Fh1,k (z), |x| ≤ w1 2, 0 ≤ z ≤ l1 , T2k Jh1 (x, z) = Ah1k w1 π 2 w 1 k=0     M   x 2 4 − 2 · δk0  Fh1,k (z), |x| ≤ w1 2, 0 ≤ z ≤ l1 , Jh1 (x, z) = Ah1k T2k w1 π w1 2 k=0 (13) 2  2k 2 = χ2 − where according to (2) kzi,k = 0, Ah1(2)k – unknown coefficients of h,n wi / 2 expansion, M – the reducing order of an infinite series, Fhi,k (kz1k , z) given in the [18, 27]. It should be noted that differential Eq. (12) for the magnetic vector potential of current density Jh (x, z) with “electrical” transverse conditions is solvable when χh,n = 0. The two-dimensional function in this case written as:     √  x x sinh kx1k z 2 2 2   Jh1 (x, z) = A1,0 √ T0 A1k √ T2k , ·z+ π w1 π w1 w1 2 w1 2 kx1k cosh kx1k l1 k=1



for |x| ≤ w1 2, 0 ≤ z ≤ l1 and     √  x x sinh kx2k (L − z) 2 2 2   Jh2 (x, z) = A2,0 √ T0 A2k √ T2k · (L − z) + π w2 π w2 w2 2 w2 2 kx2k cosh kx2k l2 k=1



for |x| ≤ w2 2, l1 ≤ z ≤ L, and kxi,k = w2k , i = 1, 2. The current component with i/ 2 χh,n = 0 has physical meaning of the constant component of longitudinal current Jz . Similarly, the two-dimensional function for the electric vector potential of the current density in partial regions 1 and 2 written as series:    M   x 2 2  Fe1,k (z), |x| ≤ w1 2, 0 ≤ z ≤ l1 , Je1 (x, z) = Ae1k √ U2k+1 π w1 w1 2 k=0 Je2 (x, z) =

M  k=0



2 Ae2k √ π

   x 2  Fe2,k (z), |x| ≤ w2 2, l1 ≤ z ≤ L, U2k+1 w2 w2 2

(14)

368

Y. Rassokhina et al.

2 = χ2 − (2k+2) , A where kzi,k e1(2)k are unknown coefficients of series expansion and e,n (wi / 2)2 Fe1,k (kz1k , z) given in [18, 27]. Thus, obtain a linear system of equations for unknown coefficients of expansion from the condition of continuity of functions at the boundary of partial regions. The equation for spectrum of eigenvalues χ2h(e),m , m = 1, 2 . . . is a    condition for solution a homogeneous system of equations detαkn χh(e),m  = 0, see algorithm details in [27]. Solution of the “electric” and “magnetic” boundary value problems for eigenfunctions of current density in an irregular shielded strip line is used to algebraization boundary value problems for resonators with discontinuities containing inductive or capacitive m m sections by the formula (9). The coupling integrals αm h,q,mn , βh,q,mn and γh,q,mn between the inhomogeneous strip resonator and the volume rectangular resonator are calculated by the formulas:  

m dS, β = ∇J = ∇Jh,q (x, z)∇φmn (x, z)dS αm z) ∇ψ z) × e mn (x, y h,q (x, h,q,mn h,q,mn 2

SMSL

 m = γh,q,mn

SMSL





∇Je,q (x, z) × ey · ∇ψmn (x, z) × ey dS.

SMSL

where φmn (x, z), ψmn (x, z) are the basis functions of the electric and magnetic vector potential of a three-dimensional rectangular resonator (6). Figure 3 shows the solution of “electric” (e.w.–e.w.) and “magnetic” (m.w.–m.w.) boundary-value problems for eigenvalues χh(e),m , m = 1, 2 . . . of vector potentials of magnetic and electrical type   for a band resonator with a capacitive segment in it (Fig. 2) at a ratio a = w2 w1 = 1 3. The results were obtained by limiting the series (13), (14) to the values of M = 3 for both boundary value problems. After increasing the reducing order M, the values of the eigenvalues deviate from the value obtained at M = 3 no more than 10−3 .

Fig. 3. Eigenvalues χh,n and χe,n of the basis functions of magnetic J h (x, z) and electric J e (x, z) vector potentials of a microstrip resonator with capacitive section obtained from the solution of “electric” and “magnetic” boundary value problems (parameter a = 1/3).

The Analysis of Distributed Two-Layers Components

369

The results of the convergence investigation of the algorithms for solving “electrical” and “magnetic” boundary problems relative to the resonant frequencies of a microstrip resonator with capacitive discontinuity are presented in [27]. As example, for the structure with parameters: substrate material of the thickness h = 0.635 mm with a dielectric constant εr = 10.2, the width and height of the screen of the three-dimensional resonator are A = 15.0 and b1 = 7.0 mm, respectively, other structure parameters (in mm): w1 = 2.4, w2 = 0.58 (characteristic impedance of the main transmission line is Z 0 = 50 Ohms), l1 = 2.2 (l = 4.4), l 2 = 15.125, the calculations of resonator eigenfrequency can be carried out with the parameters N = 300, M = 4, P = 4 reducing of a series (6), (9) (13) and (14). By using of orthogonal polynomials basis to describe the current density the value of M reduces to M = 3 ÷ 4, and with it increasing an effect of numerical instability is also observed [22]. Figure 4 shows the characteristics of the reflection and transmission coefficients of the main  wave of a microstrip line on inductive discontinuities for three values of a = w1 w2 . The main transmissionline has a width w2 = 2.62 mm, and the lengths of the inductive section are equal to λ 16 at frequency of 3.0 GHz (for example, section of such length are used for the design of short transitions and filters [23–25]).

Fig. 4. (a) Eigenfrequency spectrum obtained from solution of boundary value problems for inductive discontinuity in a microstrip transmission line (Fig. 6); (b) scattering characteristics of inductive section in the microstrip transmission line (a = w1 /w2 ). Screen and substrate parameters for calculations (in mm): h = 0.635, εr = 10.2, A = 15.0, b1 = 7.0

Figure 5a shows the eigen frequency spectrum obtained from the solution of the “elec tric” and “magnetic” boundary value problems in dependence of the ratio a = w2 w1 of the width of the microstrip lines at a fixed width w1 = 0.58 mm  and l 2 = 1.2 mm (the length of the section l = 2l1 = 2.4 mm corresponds to the length λ 16 at 3.0 GHz), and Fig. 5b shows characteristics of the reflection and transmission coefficients of the capacitive discontinuity in the microstrip transmission line obtained from it by the TRT. It can be seen that the scattering characteristic of capacitance discontinuity in the microstrip transmission line over a wide frequency range  (from 1.5 to 6.0 GHz) is smooth, and its slope of curve depends on the ratio a = w2 w1 ≤ 1: the greater difference between the linewidths leads to the smaller slope of the reflection coefficient characteristic.

370

Y. Rassokhina et al.

Fig. 5. (a) The eigen frequency spectrum obtained from solution of boundary value problems for capacitive discontinuity in the microstrip transmission line; (b) frequency dependences of the reflection and transmission coefficients on the capacitance discontinuity in dependence of the ratio a of the width of the main transmission line and the width of the capacitive section of 2·l 1 = 2.4 mm long.

Thus, algorithms for analysis of finite length of step discontinuities in a microstrip transmission line (inductive or capacitive sections) by the transverse resonance method were improved. At the same time, we used expressions of the current density in the discontinuity microstrip line due via the magnetic and electric vector potentials for the algebraization of the boundary value problem. The advantage of the proposed method is that the order of reduction of series by eigenfunctions (vector potentials) remains constant when analyzing discontinuity in a wide frequency range. The algorithms are well converged and ensure the accuracy of the calculation of the resonant frequencies in the range of 10−2 GHz, where taking into account only two or three eigenfunctions in the expansion into series of electric and magnetic vector potentials for current density is sufficient.

4 Results of Numerical Analysis of Two-Layer Discontinuities in Planar Structures The calculations were fulfilled reducing the number of basis functions (6) to 150 terms of series, polynomial basis for current density to 5 terms of series (13) and (14), P = 3−4 eigenfunctions of magnetic and electric type for current density in microstrip line (9), and 4 TE-type waveguide functions in field description (7) at the slot resonator aperture. Here and below the substrate thickness is h = 0.365 mm, dielectric permittivity is εr1 = 10.2, three-layer shield dimensions (in mm) are: A = 15 .0, B = 7.365, b2 = 5.0 (Fig. 1b). The transverse resonance technique was realized bymeans of spectral curves approximation   using the rational functions in form f (x) = 1 a0 + a1 x + a2 x2 + . . . + am xm [26]. Because of this algorithm optimization, the calculation time of discontinuity analysis is significantly reduced. In Fig. 6 shows the results of calculating the spectrum of eigen frequencies of the resonator and the corresponding scattering characteristics for the structure in Fig. 1a. As it can be seen, the scattering characteristic contains one frequency

The Analysis of Distributed Two-Layers Components

371

of the resonant transmission at 1.5 GHz and two frequencies of resonant reflection of the signal at 4.25 and 5.95 GHz.

Fig. 6. Eigenfrequencies spectrum of microstrip resonator with distributed 2-layer discontinuity in Fig. 1 obtained solving the two, “electric” and “magnetic”, boundary value problems and corresponding scattering characteristics. The structure dimensions are (in mm): w1 = 2.4, w2 = 0.58, length of capacitive section l = 2l1 = 2.4, (l1 = 1.2). The parameters of H-shaped slot resonator are: as = 8.7, bs = 5.2, s1 = 0.6, s2 = 0.6.

The Fig. 7a shows the topology of resonant structure with symmetric 2-layer discontinuity in 3-layer planar structure consisting of a step width in a microstrip transmission line from w1 to w2 , (capacitive section of a microstrip line) and a comb-shaped slot resonator in its ground plane. The Fig. 7b shows their characteristics of the transmission and reflection coefficients calculated by the transverse resonance technique (TRT). The measured scattering characteristics for the experimental prototype are shown in Fig. 7c. As it can be seen, the device with a discontinuity of three directly coupled slot resonators provides attenuation in the wide frequency band from 4.5 to 8.8 GHz and has the bandwidth from 3.3 to 3.6 GHz. With regard to the useful frequency characteristics, the so-called dumbbell-shaped slots [28] are in the area of interest among the known relatively simple DGS configurations. In the case of a series of such slots [29] microstrip line has clearly separated bandwidth and wide stop band in the transmission characteristics, ensures unwanted frequencies cut-off. Such a configuration of discontinuities in the microstrip line performs the function of a low-pass filter (LPF) having more compact physical size in comparison with conventional filters [30]. We consider also topology of the planar structure shown in Fig. 8. The structure is symmetric, it contains a low-impedance insertion of the microstrip line, under which there are two identical “dumbbells” with rectangular heads – slots etched in the ground layer of substrate.

372

Y. Rassokhina et al.

Fig. 7. (a) Topology and (b) scattering characteristics of a two-layer distributed discontinuity; (c) experimental prototype. Parameters of structure (in mm): substrate h = 0.635, εr = 10.2, ideal metallic shield A = 15.0, b1 = b2 = 7.0, microstrip w1 = 2.7, w2 = 1.2, lw1 = 1.2, comb slot resonator ls1 = 9.15, ls2 = 5.1, bs = 3.6, s = 0.3, s1 = s2 = 0.4.

Based on the topology structure, in [31] we proposed a strategy of step-by-step optimization of filter geometry together with proportional ratios of the geometric parameters for initial approximation, which provide the transfer function of the low-pass filter. The

The Analysis of Distributed Two-Layers Components

373

Fig. 8. Considered topology of the microstrip line: 1 – substrate, 2 - microstrip line, 3 – etched dumbbell-shape slots, 4 – ground layer.

strategy allows to take into account the complex reflection coefficient at the fundamental frequency and higher harmonics and in the case of using this topology as a main element of load network of the power amplifier allows to approach the desired conditions of a particular class of amplifier. For purposes of easy reproducibility of topology and convenient parametric geometry description, we linked the geometric parameters of the structure to a character size x 0 . For the LPF targeted at fundamental frequency f 0 = 2.155 GHz, the characteristic size is x 0 = 20.4 mm (see Fig. 8), the low impedance segment of the microstrip line has a length l = x 0 /2 and a width w1 = x 0 /6 whose impedance is Z = 16.5 Ohms. A pair of dumbbells in the ground layer has the same dimensions g 1 = w, d = x20 , a = b = d −w 2 and the distance between vertical symmetry axes of slits is limited by constraints a < s = x40 < l, where width of the microstrip line w = 0.58 mm corresponds to the impedance Z 0 = 50 Ohms. There are some approaches to construct equivalent circuits of the structure under consideration [30], but with such compactness theelements of topology and the space gaps between them have sizes comparable to λg 4, interact with each other and act as a single structure with a complex pattern of scattering. Since empirically chosen parameters of the equivalent schemes, it does not give advantages in the prediction of filter properties. The influence of the geometric parameters l0 , x 0 , w1 , d, l, s (see Fig. 8) on the filtering properties was investigated with constraints (s > a, s + a < l) in [31]. Figure 9a shows dependency of scattering characteristics from w1 at fixed rest parameters. According to the numerical analysis results, two variants of the topology after the optimization were selected, according to which samples of LPF with planar dumbbellDGS structures on substrate (H = 0.64 mm, εr = 10.2, tan δ = 0.0022) were made for experimental measurements [32]. Measurement shows that at least at the working frequency and second harmonic studied samples behave as expected (see Fig. 9b). Cutoff 3 dB is located between the main and second harmonics, the ratio f 0 /f c ≈ 0.78, where f c and f 0 are the cutoff frequency and resonance, respectively. Configuration with chosen substrate allows to guarantee a narrow bandwidth near the fundamental frequency within the band from ≈1.5 to ≈2.5 GHz.

374

Y. Rassokhina et al.

Fig. 9. (a) Simulated scattering characteristics of three-layer planar structure with twin dumbbell shaped slots and variation of width w1 of low-impedance microstrip line extension from λ 10 to  λ 4; (b) comparison of measured and simulated scattering characteristics of a designed sample.

5 Conclusion The technique of analyzing two-layer distributed discontinuities in three-layer planar structures by the method of transverse resonance is proposed and implemented. Methods of algebraization of boundary value problems for basic discontinuities used in such structures are proposed in the form of a step discontinuity in a microstrip transmission line (capacitive or inductive section) and slot resonators of complex shape in its ground plane. Algorithms for the solution of boundary value problems are constructed to calculate its eigen (resonant) frequencies of a volume resonator, from which the scattering characteristics of discontinuities are calculated. It has been demonstrated that current density in nonhomogeneous strip line is well described by the magnetic and electric vector potential function, which is expressed as decomposition on a Chebyshev polynomials basis. The field in the slot resonator, in particular comb-slot resonator, is written as series of eigenfunctions Th(e)y,k of the TEand TH-waves of complex-shaped waveguide. Both methods of algebraization of the boundary value problem take into account the physical singularities of the field behavior on the discontinuity components, provide the system of linear algebraic equations of small order for the calculation of the spectrum of eigen frequencies and fast convergence of the algorithm. The complex-shaped slot resonator, symmetrically located in the microstrip transmission line ground plane, provides broadband rejection of higher harmonics of the fundamental frequency, and in combination with stepped discontinuity in microstrip line, it can be used to design multifunctional devices having resonance reflection and resonance transmission frequencies. Considered low-pass filter on symmetrical twin dumbbell-shaped slots in the microstrip transmission line ground plane is characterized by compact size, wide stop band at higher harmonics and the desired frequency dependence of the impedance hodograph, which is a prerequisite for the construction of a high-power microwave power amplifier.

The Analysis of Distributed Two-Layers Components

375

References 1. Jansen, R.H.: The spectral-domain approach for microwave integrated circuits. IEEE Trans. Microw. Theory Tech. 33(10), 1043–1054 (1985) 2. Kewano, K.: Hybrid-mode analysis of coupled microstrip-slot resonators. IEEE Trans. Microw. Theory Tech. 33(1), 38–43 (1985) 3. Schwab, W., Menzel, W.: On the design of planar microwave components using multilayer structures. IEEE Trans. Microw. Theory Tech. 40(1), 67–72 (1992) 4. Itoh, T., Hebert, A.S.: A generalized spectral domain analysis for coupled suspended microstrip lines with tuning septum’s. IEEE Trans. Microw. Theory Tech. 26(10), 820–826 (1978) 5. Itoh, T.: Spectral-domain immittance approach for dispersion characteristics of generalized printed transmission fine. IEEE Trans. Microw. Theory Tech. 28(7), 733–736 (1980) 6. Fukuoka, Y., Neikirk, D.P., Itoh, T.: Analysis of multilayer interconnection lines for a highspeed digital integrated circuit. IEEE Trans. Microw. Theory Tech. 33(6), 527–532 (1985) 7. Kim, J.P., Park, W.S.: Novel configurations of planar multilayer magic-T using microstripslotline transitions. IEEE Trans. Microw. Theory Tech. 50(7), 1683–1688 (2002) 8. Guo, X., Zhu, L., Wu, W.: Balanced wideband/dual-band BPFs on a hybrid multimode resonator with intrinsic common-mode rejection. IEEE Trans. Microw. Theory Tech. 64(7), 1997–2005 (2016) 9. Bukuru, D., Song, K., Zhang, F., Zhu, Y., Fan, M.: Compact quad-band bandpass filter using quad-mode stepped impedance resonator and multiple coupling circuits. IEEE Trans. Microw. Theory Tech. 65(3), 783–791 (2017) 10. Szydlowski, L., Jedrzejewski, A., Mrozowski, M.: A trisection filter design with negative slope of frequency-dependent crosscoupling implemented in substrate integrated waveguide (SIW). IEEE Microw. Wirel. Compon. Lett. 23(9), 456–458 (2013) 11. Zakharov, A., Rozenko, S., Ilchenko, M.: Two types of trisection bandpass filters with mixed cross-coupling. IEEE Microw. Wirel. Compon. Lett. 28(7), 585–587 (2018) 12. Zakharov, A., Rozenko, S., Litvintsev, S., Ilchenko, M.: Trisection bandpass filter with mixed cross-coupling and different paths for signal propagation. IEEE Microw. Wirel. Compon. Lett. 30(1), 12–15 (2020) 13. Itoh, T. (ed.): Numerical Techniques for Microwave and Millimeter-Wave Passive Structures. Wiley, New York (1989) 14. Bornemann, J., Arndt, F.: Transverse resonance, standing wave, and resonator formulations of the ridge waveguide eigenvalue problem and its application to the design of E-plane finned waveguide filters. IEEE Trans. Microw. Theory Tech. 38(8), 1104–1113 (1990) 15. Itoh, T., Mittra, R.A.: Technique for computing dispersion characteristics of shielded microstrip lines. IEEE Trans. Microw. Theory Tech. 22(10), 896–898 (1974) 16. Itoh, T.: Analysis of microstrip resonators. IEEE Trans. Microw. Theory Tech. 22(11), 946– 952 (1974) 17. Veselov, G.I. (ed.): Mikroelektronnye ustrojstva SVCH. Vyssh. Shkola, M. (1988). (in Russian) 18. Rassokhina, Yu.V., Kryzhanovskii, V.G.: A method for analyzing irregularities in striplineslot structures. Part 1: analysis of a width jump in a microstrip line by the transverse resonance method. Telecommun. Radio Eng. 76(8), 653–665 (2017) 19. Rassokhina, Yu.V, Krizhanovski, V.G.: Periodic structure on the slot resonators in microstrip transmission line. IEEE Trans. Microw. Theory Techn. 57(7), 1694–1699 (2009) 20. Rassokhina, Yu.V., Krizhanovski, V.G.: Analysis of coupled slot resonators of complex shape in metalization plane of a micro-strip transmission line using the transversal resonance techniques. Radioelectron. Commun. Syst. 55(5), 214–222 (2012)

376

Y. Rassokhina et al.

21. Rassokhina, Yu.V., Kryzhanovskii, V.G.: A method for analyzing irregularities in striplineslot structures. Part 2: analysis of complex irregularities in three-layer planar structures. Telecommun. Radio Eng. 76(12), 1049–1056 (2017) 22. Bornemann, J.A.: Scattering-type transverse resonance technique for the calculation of (M)MIC transmission line characteristics. IEEE Trans. Microw. Theory Tech. 39(12), 2083–2088 (1991) 23. Matthaei, G.L.: Short-step Chebyshev impedance transformers. IEEE Trans. Microw. Theory Tech. 14(8), 372–383 (1966) 24. Van der Walt, P.W.: Short-step-stub Chebyshev transformers. IEEE Trans. Microw. Theory Tech. 34(8), 863–868 (1986) 25. Rassokhina, Yu.V., Krizhanovski, V.G., Kovalenko, V.A.: Compact filters with slot resonators and fast method of its analysis. Visnyk NTUU KPI Ser. – Radiotekhnika Radioaparatobuduvannia (67), 18–24 (2016). (in Ukranian) 26. Rassokhina, Yu.V., Krizhanovski, V.G.: The analysis of distributed two-layers components in three-layer planar structure. Visn. NTUU KPI, Ser. – Radioteh. radioaparatobuduv 72, 5–12 (2018) 27. Rassokhina, Yu.V., Krizhanovski, V.G.: The microstrop step discontinuity analysis by transverse resonance technique: method of boundary value problem algebraization. Radiotekhnika: All-Ukr. Sci. Interdep. Mag. 196, 117–129 (2018). https://nure.ua/wp-content/uploads/2019/ Scinetific_editions/17.pdf. Accessed 03 Dec 2019. (in Ukranian) 28. Ahn, D., Park, D., Kim, C.S., Kim, J., Qian, Y., Itoh, T.: A design of the low–pass filter using the novel microstrip defected ground structure. IEEE Trans. Microw. Theory Tech. 49(1), 86–93 (2001) 29. Boutejdar, A., Omar, A., Burte, E.P., Mikuta, R.: An improvement of defected ground structure lowpass/bandpass filters using H–slot resonators and coupling matrix method. J. Microw. Optoelectron. Electromagn. Appl. 10(2), 295–307 (2011) 30. Khandelwal, M.K., Kanaujia, B.K., Kumar, S.: Defected ground structure: fundamentals, analysis, and applications in modern wireless trends. Int. J. Antennas Propag. 2017, 22 p. (2017). Article ID 2018527 31. Komarov, V., Barybin, O., Rassokhina, Yu.V., Krizhanovski, V.G.: Dumbbell-shaped defected ground structure resonator filter for high-efficiency microwave power amplifiers. In: 2018 International Conference on Information and Telecommunication Technologies and Radio Electronics (UkrMiCo), pp. 1–4, Odessa (2018). https://doi.org/10.1109/ukrmico43733.2018. 9047537 32. Komarov, V., Rassokhina, Yu.V., Krizhanovski, V.G.: Synthesis of the compact low pass filter using dumbbell-shaped slot resonators. Radiotekhnika 197, 50–55 (2019). (in Ukranian)

Planar Bandpass Filters with Mixed Couplings Alexander Zakharov(B)

, Mykhailo Ilchenko(B) and Ludmila Pinchuk(B)

, Sergii Rozenko(B)

,

National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Peremoga ave. 37, Kyiv 03056, Ukraine [email protected], [email protected], [email protected], [email protected]

Abstract. The problem so designing stripline and microstrip bandpass filters with mixed coupling, including the magnetic and electric components of the interaction, are considered. It is shown that the transmission zero corresponding to mixed coupling coefficients can be shifted long the frequency axis by changing the shape of the stepped-impedance resonators. It is confirmed that N-resonator planar filters can have (N – 1) transmission zeros. Designs of microstrip filters with combined coupling, which include mixed coupling and the traditionally used magnetic and electric coupling, are proposed. It is shown that the number of transmission zeros of such filters is smaller than for filters with only mixed coupling, but their designing and tuning are less labor-consuming. The data of the experiment and computer simulation are presented. Keywords: Microstrip bandpass filter · Mixed coupling coefficient · Cross-coupling · Transmission zero · Frequency response

1 Introduction Stripline [1–3] and microstrip [4, 5] bandpass filters are used in the transceiver equipment. For most filters, increased selectivity is required. One of the ways to in crease the filter’s selectivity is to in clued in its frequency response transmission zeros located the left and right of the passband. Transmission zeros can be implemented by means of cross-coupling filters [6–8]. More recently, a new approach to introducing transmission zeros began to develop; it is based on mixed coupling between adjacent resonators [9–15]. Mixed-coupling coefficients k can take both positive and negative values. This promising approach was used for filters of various types: coaxial [9], multilayer [10], and microstrip [11–15]. However, we are not aware of any work in which this approach is applied to strip bandpass filters. It should also be noted that, in microstrip filters, there is a fairly strong electromagnetic interaction between non-adjacent resonators, which significantly affects the filter’s characteristics. In the above-cited works, this fact was not taken in to account. If we simulate the frequency response of an idealized filter with only mixed coupling and then “connect” weak cross-coupling, then the original response of the filter will be degraded. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 377–393, 2021. https://doi.org/10.1007/978-3-030-58359-0_21

378

A. Zakharov et al.

In this paper, we consider the control of transmission zeros and implementation of elliptic frequency response in strip and microstrip bandpass filters with mixed coupling. The cross-coupling in microstrip filters is taken in to account and suppressed.

2 Identification of Coupling Frequencies The coefficient of electromagnetic coupling between two identical resonators can be calculated by the formula [4] k=

fo2 − fe2 fo2 + fe2

(1)

where f e and f o are the frequencies of the even and odd modes, respectively (they are also called the coupling frequencies). If f o > f e , then k > 0, and, if f o < f e , then k < 0. The values of k are expressed both in units and percents. Figure 1a shows a pair of quarter-wave resonators with a characteristic impedance Z 0 and an electric length θ, which are coupled by a lumped LC circuit. At there sonant frequency f 0 , we have θ = π/2. The capacitance C 1 is very small; it provides a weak coupling of resonators with loads. In the absence of inductance L, the coupling between the resonators is capacitive, characterized by the coupling coefficient k C . If we remove the capacitance, then the coupling will be inductive, characterized by the coefficient k L . In general, the coupling is mixed, in which both inductive and capacitive components are present. Obviously, the resonance frequency of parallel LC circuit, √ (2) fz = 1/2π LC, is a transmission zero, because, at this frequency, no transfer of energy from one resonator to another takes place. Transmission zero are also termed “attenuation poles”. We will be interested in the relative position of the frequencies f z and f 0 , as well as the identification of the coupling frequencies f e and f o with respect to the frequency response of the circuit under consideration. The coupling coefficient between identical parallel-type resonators can be determined from the general expression [1] k = J /b,

(3)

where b is the steepness of the resonator’s conductivity and J is the parameter of the conductivity inverter, which replaces the coupling circuit. If the resonators are connected only through an inductance, then J = 1/2πf L. Substituting this expression into formula (3) gives the inductive coupling coefficient at the resonant frequency: kL ≈ 1/2π f0 Lb.

(4)

If the resonators are connected only through a capacitance, then J = 2πfC and kC ≈ −2π f0 C/b.

(5)

Planar Bandpass Filters with Mixed Couplings

379

Fig. 1. Identification of coupling frequencies of two quarter-wave resonators for different types of coupling between them: (a) schematic view; (b) k = k L ; (c) k = k C ; mixed coupling with (d) k > 0, (e) k < 0, and (f) k = 0; (solid curves) |S 21 | and (dashed curves) phase of S 21 .

The capacitance coupling coefficient in expression (5) has the minus sign. If the resonators are connected through an LC circuit, then k ≈ kL + kC ,

(6)

where the coupling coefficients k L and k C are defined by expressions (4) and (5). Expressions (4)–(6) can be applied to any pair of identical resonators, including quarter-wave and half-wave. At the same time, they are approximate, because the real coupling elements are replaced by an inverter. Unlike the inverter, the coupling elements C and L shift the resonant frequency of the resonators. The smaller this shift, the more accurate formulas (4)–(6).

380

A. Zakharov et al.

By choosing the values of L and C in (2), one can satisfy the conditions f z < f 0 and f z > f 0 , under which a transmission zero is located below or above the passband. Let us find the signs of k with which such apposition is possible. The mixed coupling coefficient k is negative if k L < |k C | and, positive, if k L > |k C |. The first condition can be represented in the form 1/2π f0 Lb < 2π f0 C/b or 1/(2π)2 LC < f02 . Since the product LC determines transmission zero (2), we have fz < f0 for k < 0.

(7)

The second condition, k L > |k C |, leads us to the inequality fz > f0 for k > 0.

(8)

In equalities (7) and (8) show that, if the coupling between two resonators is mixed, then it is impossible to place f z below f 0 if k > 0, and impossible to place f z above f 0 if k < 0. In expressions (4) and (5), f can be different. Let us find the values of k C and k L at the zero transmission, f 0 = f z . Taking into account (2), we obtain  1 C . kL (fz ) = |kC (fz )| = b L At the frequency f z , the coefficients k L and k C have the same absolute value and opposite signs; therefore, k(fz ) = 0.

(9)

Figures 1b–1f shows the transmission frequency response (the absolute value and the phase of S 21 ) of the considered pair of resonators weakly coupled with the loads. In the construction, it was assumed that the resonators are quarter-wave at the frequency f 0 = 1 GHz and Z 0 = 10  and C 1 = 0.01 pF. Expressions (4) and (5) include the steepness parameter of the input susceptance of the resonators at the connection point of the coupling elements, b. If the coupling elements are connected to the open ends of the quarter-wave resonators, then b = π/4Z 0 [1]. This equality and expressions (4) and (5) allow one to find the values of L and C from a given value of k: kL L, nH kC C, pF 0.02 101.32 −0.02 0.25 0.04 50.66 −0.04 0.5 0.06 33.77 −0.06 0.75 0.08 25.33 −0.08 1.0

Planar Bandpass Filters with Mixed Couplings

381

Figure 1b corresponds to purely inductive coupling with k L = 0.02. The coupling frequencies f e and f o can be distinguished by the phase response. At these frequencies, the phase changes step-wise. If the phase of S 21 , changing from left to right, crosses the lowest coupling frequency and increases step-wise, then this frequency corresponds to an even oscillation mode, f e , and the coupling coefficient is positive. If, under the same conditions, the phase of S 21 decreases step-wise, then the lowest coupling frequency corresponds to an odd oscillation mode, f o , and the coupling coefficient is negative (see Fig. 1c for a purely capacitive coupling with k C = – 0.02). The same regularity takes place for a mixed coupling, in which k can be either positive or negative (see Fig. 1d and 1e). In these cases, we have an additional phase jump appears at the zero transmission frequency, which is located on the side of the odd oscillation mode, f o . With a mixed coupling, the location of the transmission zero f z also allows us to distinguish between the frequencies f o and f e . In Fig. 1f, the case of a mixed coupling with k = 0 (f z = f 0 ) isreflected. It should be noted that the regularities shown in Fig. 1 take place when the input and output loads are connected to the resonators in-phase. If we use an anti-phase connection of loads, as shown in Fig. 2a for the case of half-wave resonators, the phase response changes to the opposite one. For half-wave resonators at a resonance frequency f 0 = 1 GHz, we have θ = π, Z 0 = 10 , and a steepness parameter b = π/2Z 0 [1]. Figure 2b shows the case of an inductive coupling with k L = 0.02, and Fig. 2c, the case of a mixed coupling with k = 0.02. The phase jump at the frequency f z also takes place in the case of antiphase connection of loads.

Fig. 2. Coupling frequencies of two half-wave resonators with antiphase connection of loads: (a) schematic view; (b) k = kL; (c) mixed coupling, k > 0; (solid curves) |S21| and (dashed curves) phase of S21.

382

A. Zakharov et al.

Thus, the coupling frequencies can be distinguished by the phase response. In this case, it is necessary to take into account whether the connection of loads in-phase or antiphase. The presence of a transmission zero f z in the case of a mixed coupling facilitates the identification of frequencies f o and f e . The odd oscillation mode, f o , is nearest to f z , which allows one to identify the coupling frequencies without addressing to the phase response. By varying the L and C of the coupling circuit, one can move transmission zero f z (2) along the frequency axis. Moreover, f z can be moved at a given value of the mixed coupling coefficient k. This follows from formula (6) and the ratio k L /k C = f 2z /f 20 , which is obtained from expressions (4), (5), and (2). The last expression can be written in a different form:  fz = f0 kL /kC . (10) The variation in the ratio k L /k C of the coupling coefficients in (10) while preserving their difference makes it possible to shift the transmission zero at a given value of the mixed coupling coefficient (see Fig. 3 for the pair of resonators shown in Fig. 1a). The closer the ratio k L /k C to unity, the closer the transmission zero to f 0 .

Fig. 3. Displacement of transmission zero with mixed coupling coefficient |k| = 0.02: (1) kL/kC = 2(fz = 1.414 GHz); (2) kL/kC = 3/2 (fz = 1.225 GHz); (3) kL/kC = 2/3 (fz = 0.816 GHz); (4) kL/kC = 1/2 (fz = 0.707 GHz); (solid curves) k > 0, (dashed curves) k < 0.

3 Pairs of Resonators with Mixed Coupling Figure 4 shows some pairs of microstrip and stripline resonators with mixed coupling. The electromagnetic interaction between the resonators has the magnetic and electrical components with the coefficients k L and k C , respectively. If the dominating component is magnetic, then k > 0, and, if electric, then k < 0. The resonators are placed close to each other, since the electric interaction acts only at small distances.

Planar Bandpass Filters with Mixed Couplings

383

Fig. 4. Mixed electromagnetic coupling between two microstrip or stripline resonators: (a) quarter-wave; (b) half-wave; (c) quarter-wave with two coupling strips; (d) half-wave with two coupling strips; (e) quarter-wave with a moving coupling strip; (f) half- wave with a coupling strip; (g) stepped-impedance and quarter-wave resonators with a coupling strip; (h) quarter-wave resonators with a common short-circuited end.

The pairs of microstrip resonators shown in Figs. 4a and 4b were studied in [5, 16]; the geometric parameters in the figure (in mm) do not need explanation. The same pairs of stripline resonators of the structure were considered in [17, 18]. In the pairs of resonators shown in Figs. 4c and 4d, the coupling is controlled by two strip lines. Reducing the gap between the stripline and the open ends of the resonators increases the electric component of the coupling [19]. An increase in the strip width near the short-circuited ends of the resonators increases the magnetic interaction [20]. The strip in Fig. 4d is located strictly in the middle of the resonators; an increase in its width also increases the magnetic coupling. A pair of quarter-wave resonators with a strip moving along the resonators (Fig. 4e) was considered in [14]. An increase in the distance between the strip and the short-circuited ends of the resonators increases the magnetic interaction. A similar effect takes place with an increase in the distance between the strip and the middle of the half-wave resonators (Fig. 4e) [21]. A pair including a quarter-wave and stepped-impedance resonator (Fig. 4g) may be useful in some filter designs [22]. A pair of microstrip resonators (Fig. 4h) was considered in [15]. Two-cavity bandpass filters based on pairs of resonators shown in Fig. 4, which have an a symmetric frequency response, are useful in some applications. For example, Fig. 5 shows the measured frequency response of two microstrip filters on substrates with εr = 92 and a thickness h = 1 mm. The inserts show the topologies of these filters, which occupy an area of 10 mm × 6 mm and contain stepped-impedance half-wave resonators

384

A. Zakharov et al.

with a central conductor width of 2 mm. In both filters, the transmission zero is located at a distance of 60 MHz from the edge of the passband measured in the absolute attenuation level of 2.5 dB.

Fig. 5. Measured frequency response of microstrip two-resonator filters with mixed coupling for (a) k > 0 and (b) k < 0.

It is extremely important to control the position of the transmission zero f z at a fixed mixed coupling coefficient k. Expressions (4) and (5) make it possible to calculate the values of k L and k C in the case of using an LC circuit as a coupler. The variation in the ratio k L /k C (10) while preserving the difference between these quantities solves the problem of moving f z relatively easily: by calculation. Apparently, this problem for electromagnetically coupled microstrip resonators can also be solved on the basis of the energy approach [16]. Computer simulation makes it possible to detail the process of moving the transmission zero f z in pairs of microstrip and strip resonators with mixed coupling. Figure 6 shows the frequency characteristics of different modifications of a pair of quarter-wave microstrip stepped-impedance resonators (see Fig. 4a). They used substrates with εr = 10 and a thickness h = 2 mm. Short circuiting was performed by metalized holes. In the simulation by the Microwave Office (AWR) software, we used w1 = 3.6 mm, the area occupied by the resonators of 20 mm × 16 mm, and the height of the metal screen above the substrate of 20 mm. The values of S, L, and l varied in such a way that the two peaks on the transmission response coincided with the frequencies of 1.62 and 1.69 GHz. These positions of the coupling frequencies, according to (1), corresponds to the coupling coefficient |k| = 0.0423. In the insets to Fig. 6a and 6b, three pairs of resonators in which the specified value of |k| is realized are shown. The gap width S in each of the three pairs decreases from left to right, taking successively the values of 0.9, 0.6, and 0.2 mm. In the pairs of resonators with k < 0 (Fig. 6a), the width of the narrow part of the resonators is w2 = 1.8 mm. With a decrease in the gap width S between the resonators and a simultaneous decrease in the ratio l/L, the transmission zero f z approaches f 0 from the left. In the pairs of resonators with k > 0 (Fig. 6b), w2 = 3 mm. A decrease in S and l/L also approaches f z and f 0 but from the right side. Figure 7 shows the dependence of f z /f 0 on the gap width between the resonators. As S decreases, the ratio f z /f 0 tends to unity. This regularity can be explained as follows.

Planar Bandpass Filters with Mixed Couplings

385

Fig. 6. Displacement of transmission zero fz in a pair of quarter-wave microstrip resonators with a mixed coupling for (a) k < 0 and (b) k > 0 for S = (1) 0.9, (2) 0.6, and (3) 0.2 mm and different ratios (a) (1)l/L = 0.496 and fz/f 0 = 0.815, (2) l/L = 0.214 and fz/f 0 = 0.878, (3) l/L = 0.055 and fz/f 0 = 0.921; (b) (1) l/L = 0.224 and fz/f 0 = 1.164, (2) l/L = 0.159 and fz/f 0 = 1.127, (3) l/L = 0.063 and fz/f 0 = 1.077.

Since the difference between k L and k C is fixed (|k| = const), according to (10), the simultaneous increase in k L and k C approaches the ratio k L /k C to unity and narrows the distance between the frequencies f z and f 0 . The increase in k L and k C is achieved by reducing the gap width between the resonators.

Fig. 7. fz/f 0 vs. gap width between resonators for (1) k > 0 and (2) k < 0.

4 Construction of Stripline Filters with Mixed Coupling An essential feature of filtering structures based on symmetric strip transmission lines is the absence of electromagnetic interaction between non-adjacent resonators. This allows one to place the resonators one near another along a single line, forming the so-called inline filters [13], and avoiding cross-coupling. The structure of such bandpass stripline filters can be a comb or an array. When constructing them, it is sufficient to take into account only the coupling between adjacent resonators [1]:   ki,i+1  = √FBW gi gi+1

(11)

386

A. Zakharov et al.

and the external Q-factors of the end resonators, Qe1 =

g0 g1 , FBW

Qe2 =

gn gn+1 , FBW

(12)

where FBW is the relative bandwidth and gi are the parameters of the low-frequency prototype. The calculation procedure is standard with the only difference that the lefthand side of expression (11) uses the absolute value of the coupling coefficients, since their signs are different. At the initial stage of constructing, using the initial data and formulas (11) and (12), we can determine the coupling coefficients and external Q-factors of the end resonators. Each value of the mixed coupling coefficient k i is accompanied by the transmission zero f zi . A bandpass N-resonator filter with mixed coupling has (N – 1) coupling coefficients k i and the same number of transmission zeros f zi . The values of f zi are specified as the initial data. At the next stage of constructing the filter, which is most labor-consuming, for a given pair of k i and f zi , the configuration of the resonators and the gap width between them are determined (see Figs. 6 and 7). At the final stage, the connection of the end resonators to the loads that provides the required external Q-factors obtained in the first stage is determined. Figure 8 shows the result of constructing axis-resonator stripline comb filter and its simulated frequency response for the case of negligible losses. The dimensions of the filter are 13 mm × 6.5 mm × 2 mm, and εr = 92. The filter has five transmission zeros and five coupling coefficients between resonators, three negative and two positive.

Fig. 8. Comb stripline filter with mixed coupling: (a) topology (dimensions in mm); (b) simulated frequency response.

Thus, N-stripline-resonator bandpass filters with mixed coupling can have (N – 1) transmission zeros. Moreover, these filters are not nonminimal-phase, and they are not cross-coupled. In such filters, it is possible to realize an elliptic frequency response.

5 Microstrip Filters with Mixed Coupling Comb line and array type bandpass filters with quarter-wave and half-wave resonators, respectively, are most compact. In contrast to stripline filters, the magnetic interaction

Planar Bandpass Filters with Mixed Couplings

387

between non-adjacent resonators in microstrip filters is rather strong. The thicker the dielectric substrate and the closer the non-adjacent resonators, the stronger the interaction. Because of cross-coupling, such filters “involuntarily” have a non-minimal phase with a steeper right-hand slope of the frequency response [23–26]. To weaken this interaction, a special arrangement of resonator is used. In filters with half-wave resonators connected by capacitive gaps [1], the resonators are oriented with their ends towards one another. The length of such filters is large. In filters with parallel coupled half-wave resonators, adjacent resonators are displaced relatively to each other by half their length [1]. The nearest non-adjacent resonators are displaced relatively to each other by an entire length, which substantially reduces the interaction between them. For filters with U-shaped half-wave resonators [27–30], the distance between non-adjacent resonators is increased due to the sufficiently large width of such resonators. The influence of cross-coupling on the shape of the frequency response is considered on the example of a three-resonator comb filter (Fig. 9a). The magnetic cross-coupling is represented by the inductance L 13 . The filter will be assumed to be symmetric: k 12 = k 23 and Qe1 = Qe2 = Qe . The parameters of the resonators are taken the same as in Fig. 1a. When constructing the transfer characteristics of the filter, it is necessary to specify the unloaded Q-factor of the end resonators: Qe = RL b = RL π/4Z0 cos2θ  , 

(13)

where θ is the coordinate of the conductive connection of the input and output loads and RL = 50 . As is known, in the case of a three-resonator filter with ordinary (not mixed) main coupling, the character of the frequency response depends only on the sign and the magnitude of the cross-coupling. If k 13 > 0, then the steep slope is on the right, and, if k 13 < 0, on the left. If k 13 = 0, the frequency response cannot be symmetric. These positions are illustrated in Fig. 9b for a purely capacitive coupling: k 12 = k 23 = k = –4% (the inductances L 12 and L 23 are absent). The external Q-factor of the end resonators is taken equal to Qe = 25.72, which is ensured by the coordinate of connection  of the loads, θ = 67 (13). If k 13 = 0, then the frequency response is close to symmetric in the attenuation region up to 40 dB (curve 1). As k 13 increases, the steepness of the right slope increases (curves 2 and 3). In Fig. 9c, the case of a mixed coupling between adjacent resonators for which k 12 = k 23 = k = –4% (k L /k C = 1/2) is considered. If k 13 = 0, the left slope of the frequency response is steep and the transmission zero is found at the frequency f z = 0.707 GHz (curve 1). For k 13 = 0.6%, the frequency response becomes close to symmetric (curve 2). A further increase in the cross- coupling to k 13 = 1.2% increases the steepness of the right slope of the frequency response (curve 3) with transmission zero at the frequency f z = 1.148 GHz. The cross-coupling has significantly changed the form of the frequency response. The transmission zero to the left of the passband has moved to the right side, and, in the intermediate state, the frequency response had a symmetric shape. Thus, the cross-coupling “destroys” the original frequency response of a filter with mixed coupling and its influence should be reduced. The question arises whether it is possible to construct N-microstrip resonator filters with mixed coupling that would have (N – 1) transmission zeros located on both sides of

388

A. Zakharov et al.

Fig. 9. Influence of cross magnetic coupling on the frequency response of a three-resonator comb filter: (a) circuit; (b) frequency response with unmixed negative coupling between resonators, k 12 = k 23 = –4%, (1) k 13 = 0 (f z is absent), (2) k 13 = 0.6% (f z = 1.266 GHz), (3) k 13 = 1.2% (f z = 1.096 GHz); (c) frequency response with mixed negative coupling between resonators, k = –4%, (1) k13 = 0 (fz = 0.707 GHz), (2) k13 = 0.6% (fz is absent), (3) k13 = 1.2% (fz = 1.148 GHz).

the passband. To solve this problem, it is necessary to minimize as much as possible the cross-coupling between resonators, which imposes strong constraints on the structure and design of the filters. For the case of three-resonator filters, the answer can be found in [14], where a microstrip filter whose topology is shown in Fig. 10a but without a strip between resonators 1 and 3 was considered. The circuit of this filter uses positive and negative mixed coupling between adjacent resonators, which are regulated by additional bridges between them. In their absence, k 12 and k 23 < 0. Below we shall use the notation presented in Fig. 4a. The maximum effect of the attenuation of the cross-coupling between the first and third resonators takes place at a ratio of their lengths l/L ≈ 1/2 and an increase in the ratio w1 /w2 . A further decrease in k 13 results from a decrease in the substrate thickness h. In [14], the following characteristic parameters of the resonators were used: h = 0.508 mm, εr = 2.2, w1 = 1.7 mm, and w2 = 0.4 mm (w1 /w2 = 4.25). The effect of decreasing k 13 was achieved due to a very thin short-circuited segment of the stepped-impedance resonator (w2 = 0.4 mm) and a thin substrate (h = 0.508 mm). In [13], stepped-impedance half-wave resonators with extraordinary parameters w2 = 0.2 mm and w1 = 2.5 mm (w2 /w1 = 12.5) were used in order to reduce the cross-coupling. Obviously, with the parameters used in these studies,

Planar Bandpass Filters with Mixed Couplings

389

Fig. 10. Three-resonator microstrip filter with mixed coupling: (a) topology; (b) simulated frequency response; (solid curve) |S21| and (dashed curve) |S11|.

the unloaded Q-factor of the microstrip resonators Qu will be very low, which reduces the practical significance of such structures. It will be shown below that the effect of decreasing k 13 can be achieved on thick substrates and without a significant decrease in the Qu of the resonators. To do this, it is necessary to reduce the height H of the metal screen above the substrate and use an additional strip that couples resonators 1 and 3 through the capacitive gaps. For the filter in Fig. 10a, we take the following parameters: the substrate thickness h = 2 mm, εr = 10, L = 14 mm, l = 7 mm, (l/L = 1/2), w1 = 3.6 mm, w2 = 1.2 mm (w1 /w2 = 3), and the distance of 5.2 mm between resonators 1 and 3. If the middle resonator is removed and resonators 1 and 3 are quarter-wave with w = 3.6 mm, then, for H = 20 mm, we have k 13 = 3%. If the quarter-wave resonators are replaced by stepped-impedance resonators, k 13 will decrease to 0.7%. The value of k 13 is influenced by resonator 2. To measure k 13 , we reduced the length of its high-impedance segment from 7 to 3 mm so that the transmission characteristic have no three peaks and obtained k 13 = 0.9%. Reducing the height of the screen H from 20 to 4 mm reduces the cross-coupling coefficient to k 13 = 0.65%. The use of an additional strip between resonators 1 and 3 makes it possible to obtain zero coefficient k 13 and pass to the region of its negative values. In the filter shown in Fig. 10a, the gap widths between adjacent resonators are S 12 = 0.6 mm and S 23 = 1.0 mm and the mixed coupling coefficients are characterized by the signs k 12 < 0 and k 23 > 0. The area occupied by the filter is 20 mm × 19 mm. Figure 10b shows the simulated frequency characteristics of the filter for unloaded quality factors of the resonators of Qu = 250. The central frequency of the filter is f 0 = 1.59 GHz, the loss at the central frequency is IL 0 = 1.1 dB, the passband of the filter at the relative attenuation level is 1 dB BW (1 dB) = 85 MHz, and its relative bandwidth is FBW = 5.3%. Figure 11 shows a four-resonator filter with mixed coupling and its simulated frequency response. The coupling between adjacent resonators is performed by bridges and, between non-adjacent ones, by strips, like in the previous filter. The parameters of the filter in Fig. 11a are as follows: ε = 10, h = 2 mm, H = 4 mm, S 12 = 0.2 mm (k > 0), S 23 = 0.8 mm (k 23 < 0), and S 34 = 0.4 mm (k 34 > 0). The area occupied by

390

A. Zakharov et al.

the filter is 31 mm × 22 mm. In the simulation, it was assumed that Qu = 300. The central frequency of the filter is f 0 = 1.96 GHz, IL 0 = 1.3 dB, BW (1 dB) = 115 MHz, and FBW = 5.9%. The filter has two mixed couplings with the plus sign and one with the minus sign; therefore, it has two right-sided transmission zeros and one left sided.

Fig. 11. Four-cavity microstrip filter with mixed coupling: (a) topology; (b) simulated frequency response; (solid curve) |S21| and (dashed curve) |S11|.

Thus, in microstrip structures on thick substrates, it is also possible to realize bandpass filters with mixed coupling of order N with N–1 transmission zeros. In the implementation, additional means are use din order to reduce the effect of cross-coupling: a special form of resonators, their special location in the filter, strips for coupling of nonadjacent resonators, and a low-lying screen. These specificities complicate the process of designing and computer tuning of such filters.

6 Microstrip Filters with Combines Coupling The filters with combined coupling using pairs of resonators with a mixed coupling (see Fig. 4) are simpler for designing and computer tuning. In turn, these pairs a reconnected by the magnetic or electrical coupling. Denote the number of pairs of resonators by n. In such filters, the number of mixed coupling and transmission zeros is also n, the total number of resonators being N = 2n. Figures 12 and 13 show examples of microstrip four-resonator filters with combined coupling. In the filter in Fig. 12a, all resonators are aligned long a single line and the coupling between resonators 2 and 3 is magnetic. In the filter in Fig. 13a, pairs of resonators are placed at different levels and the coupling between resonators 2 and 3 is electric due to the capacitive gap. The filters use 2-mm-thick substrates with εr = 10 and the screen height is H = 20 mm. Unlike the filters in Figs. 10 and 11, they do not use resonators with a large ratio w1 /w2 , as well as additional strips for compensating the cross-coupling and the low-lying screen. The pairs of resonators in both filters in Figs. 12 and 13 are the same (see the in set to Fig. 6). In these pairs, the values of k = ±4.23% are implemented with the same gap width between the resonators

Planar Bandpass Filters with Mixed Couplings

391

Fig. 12. Four-cavity microstrip filter with magnetic and mixed coupling: (a) topology; (b) simulated frequency response; (solid curve) |S21| and (dashed curve) |S11|.

Fig. 13. Four-cavity microstrip filter with electric and mixed coupling: (a) topology; (b) simulated frequency response; (solid curve) |S21| and (dashed curve) for |S11|.

S = 0.4 mm. For the pair with k > 0, we have f z /f 0 = 1.1 and the ratio of resonator lengths is l/L = 0.203. For the pair with k < 0, we have the ratio f z /f 0 = 0.9 and l/L = 0.133. When modeling the frequency response of these filters, we assumed Qu = 300. Their main electrical parameters are the same: the central frequency f 0 = 1.646 GHz, IL 0 = 1.7 dB, BW (1 dB) = 69 MHz, and FBW = 4.2%. The area of the filter in Fig. 12 is 28.8 mm × 23 mm, and, of the filter in Fig. 13, 40 mm × 20.4 mm. The zero transmission frequencies of the filter in Fig. 12 are s lightly closer to the passband than those of the filter in Fig. 13. The filters under consideration have a small cross-coupling between the resonators. In the filter in Fig. 12, the coupling is magnetic, and, in the filter in Fig. 13, electric. This can explain the difference in the position of the transmission zeros of these two filters. The number of pairs of resonators in the filters in Figs. 12 and 13 can be

392

A. Zakharov et al.

increased. The “internal” pairs added to the circuit in Fig. 13 should contain half-wave resonators, which provide the electric coupling of this pair at two opposite ends. The considered microstrip filters with combined coupling have fewer transmission zeros than filters using only mixed coupling. However, their designing and computer tuning are less labor-consuming.

7 Conclusions Different aspects of designing planar filters with mixed coupling coefficients, in which elliptic frequency response can be implemented, have been considered. Transmission zero sin such filters are formed due to the specificities of the mixed coupling coefficients, and the cross-coupling does not participate in their formation. It has been shown that, in some cases, the effect of the displacement of transmission zeros on the frequency axis is achieved only by changing the shape of the resonators and the gap width between them. Designing microstrip filters with an elliptic frequency response requires a significant weakening of the cross coupling between the resonators. For this, additional means are needed: a special form of resonators, their special arrangement in the filter, strips for the coupling between non-adjacent resonators, and a low-lying screen. The designs of microstrip filters with combined coupling, which in clued mixed coupling, as well as magnetic or electrical coupling, are proposed. The number of resonators of such filters is even. They have fewer transmission zeros than filters with only mixed coupling. However, their designing and tuning are less labor-consuming.

References 1. Matthaei, G.L., Young, L., Jones, E.M.T.: Microwave Filters, Impedance-Matching Network, and Coupling Structures. Artech House, Norwood (1980) 2. Makimoto, M., Yamashita, S.: Microwave Resonators and Filters for Wireless Communication. Theory, Design and Application. Springer, Berlin (2001) 3. Zakharov, A.V., Il’chenko, M.E.: Thin bandpass filters containing sections of symmetric strip transmission line. J. Commun. Technol. Electron. 58(7), 728–736 (2013) 4. Hong, J.-S.: Microstrip Filters for RF/Microwave Application, 2nd edn. Wiley, Hoboken (2011) 5. Zakharov, A.V., Rozenko, S.A., Zakharova, N.A.: Microstrip bandpass filters on substrates with high permittivities. J. Commun. Technol. Electron. 57(3), 342–351 (2012) 6. Cameron, R.J.: General coupling matrix synthesis methods for Chebyshev filtering functions. IEEE Trans. Microw. Theory Tech. 47(4), 433–442 (1999) 7. Thomas, J.B.: Cross-coupling in coaxial cavity filters – a tutorial overview. IEEE Trans. Microw. Theory Tech. 51(4), 1369–1376 (2003) 8. Zhang, S., Zhu, L., Li, R.: Compact quadruplet bandpass filter based on alternative J/K inverters and λ/4 resonators. IEEE Microw. Wirel. Compon. Lett. 22(5), 224–226 (2012) 9. Wang, H., Chu, Q.-X.: An inline coaxial quasi-elliptic filter with controllable mixed electric and magnetic coupling. IEEE Trans. Microw. Theory Tech. 57(3), 667–673 (2009) 10. Tang, C.-W., You, S.-F.: Design methodologies of LTCC bandpass filters, diplexer, and triplexer with transmission zeros. IEEE Trans. Microw. Theory Tech. 54(2), 717–723 (2006)

Planar Bandpass Filters with Mixed Couplings

393

11. Ma, K., Ma, J., Member, S., Yeo, K.S., Do, M.A.: A compact size coupling controllable filter with separate electric and magnetic coupling paths. IEEE Trans. Microw. Theory Tech. 54(3), 1113–1119 (2006) 12. Chu, Q., Wang, H.: A compact open-loop filter with mixed electric and magnetic coupling. IEEE Trans. Microw. Theory Tech. 56(2), 431–439 (2008) 13. Kuo, J.-T., Hsu, C.-L., Shih, E.: Compact planar quasi-elliptic function filters with inline stepped-impedance resonators. IEEE Trans. Microw. Theory Tech. 55(8), 1747–1755 (2007) 14. Zhu, F., Hong, W., Chen, J., Wu, K.: Quarter-wavelength stepped-impedance resonator filter with mixed electric and magnetic coupling. IEEE Microw. Wirel. Compon. Lett. 24(2), 90–92 (2014) 15. Zhang, S., Zhu, L., Weerasekera, R.: Synthesis of inline mixed coupled quasi-elliptic bandpass filters based on λ/4 resonators. IEEE Trans. Microw. Theory Tech. 63(10), 3487–3493 (2015) 16. Belyaev, B.A., Titov, M.M. Tyurnev, V.V.: Izv. Vyssh. Uchebn. Zaved. Radiofiz. 8, 722 (2000) 17. Zakharov, A.V., Il’chenko, M.E., Korpach, V.N.: Features of coupling coefficients of planar stepped – impedance resonators at higher resonance frequencies and application of such resonators for suppression of spurious passbands. J. Commun. Technol. Electron. 59(6), 550– 556 (2014) 18. Zakharov, A.V., Ilchenko, M.Ye., Pinchuk, L.S.: Coupling coefficient of stepped – impedance resonators in stripline bandpass filters of array type. Radioelectron. Commun. Syst. 57(5), 217–223 (2014) 19. Zakharov, A.V., Rozenko, S.A.: Duplexer designed on the basis of microstrip filters using high dielectric constant substrates. J. Commun. Technol. Electron. 57(6), 649–655 (2012) 20. Zakharov, A.V.: Stripline combline filters on substrates designed on high-permittivity ceramic materials. J. Commun. Technol. Electron. 58(3), 265–272 (2013) 21. Zakharov, A.V., Il’chenko, M.E.: Pseudocombline bandpass filters based on half-wave resonators manufactured from sections of balanced striplines. J. Commun. Technol. Electron. 60(7), 801–807 (2015) 22. Zakharov, A.V., Ilchenko, M., Pinchuk, L.S.: Coupling coefficient of quarter-wave resonators as a function of parameters of comb stripline filters. Radioelectron. Commun. Syst. 58(6), 284–289 (2015) 23. Mao, R.-J., Tang, X.-H., Xiao, F.: Compact quarter-wavelength resonator filter using lumped coupling elements. IEEE Microw. Wirel. Compon. Lett. 17(2), 112–114 (2007) 24. Deng, P.-H., Lin, Y.-S., Wang, C.-H., Chen, C.-H.: Compact microstrip bandpass filters with good selectivity and stopband rejection. IEEE Trans. Microw. Theory Tech. 54(2), 533–539 (2007) 25. Yuceer, M.: A reconfigurable microwave combline filter. IEEE Trans. Circuits Syst. II Exp. Briefs 63(1), 84–88 (2016) 26. Brown, A.R., Rebeiz, G.M.: A varactor-tuned RF filter. IEEE Trans. Microw. Theory Tech. 48(7), 1157–1160 (2000) 27. Lin, S.-C., Lin, Y.-S., Chen, C.H.: Extended-stopband bandpass filter using both half- and quarter-wavelength resonators. IEEE Microw. Wirel. Compon. Lett. 16(1), 43–45 (2006) 28. Tang, W., Hong, J.-S.: Varactor-tuned dual-mode bandpass filters. IEEE Trans. Microw. Theory Tech. 58(8), 2213–2219 (2010) 29. Zakharov, A.V., Il’chenko, M.E.: A new approach to designing varicap-tuned filters. J. Commun. Technol. Electron. 55(12), 1424–1431 (2010) 30. Zakharov, A., Rozenko, S., Litvintsev, S., Ilchenko, M.: Trisection bandpass filters with all mixed couplings. IEEE Microw. Wirel. Compon. Lett. 29(9), 592–594 (2019)

Microstrip Monopole Antenna with Complicated Topology Dmitry Mayboroda(B)

and Sergey Pogarsky(B)

V.N. Karazin, Kharkov National University, Kharkiv 61022, Ukraine [email protected], [email protected]

Abstract. One of the possible constructions of plane-type antenna based on microstrip resonator with radiator of complex topology, excited by coaxial line segment, has been considered. The results of the study of frequency and energy characteristics of multi-band antenna based on microstrip monopole with the sectorial type inhomogeneity are presented. The main characteristics have been obtained through numerical modelling based on semi-open resonator model. The Finite Element Method (FEM) has been chosen for numerical modeling of frequency and energy parameters. The Ansoft HFSS software was used for numerical simulations. The main antenna characteristics have been analyzed. Keywords: Microstrip · Disk patch · Complex topology inhomogeneity · Radiation pattern

1 Introduction One of the current trends in microwave radio electronic equipment development is creation of universal base structures, which can be used in devices for various applications. Hence, exploration and designing of base constructions of different functional devices, such as antenna modules, mixers, filters, amplifiers, phase shifters, become important. They enable to operate both within certain frequency range and at several separate frequencies [1–8]. With regard to antenna units, complex spatial field distribution becomes required more frequently. Besides, employing new materials and modern techniques requires correct methods for calculation the functional devices parameters. Complex spatial distributions can not be formed through employing microstrip radiators of canonical topology. Making the topology more complex, insertion of additional elements, such as slot inhomogeneities of various shapes, shorting elements, etc., enable forming required spatial field distribution. This is being achieved through the change of current lines configuration on the resonator surface as well as the influence on so-called phase centers of excited oscillations. Microstrip antennas, generally, are 3D-structures. They have two-dimensional radiating patch placed on dielectric substrate. Therefore, for analysis purposes they may be considered as a 2D planar component. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 394–403, 2021. https://doi.org/10.1007/978-3-030-58359-0_22

Microstrip Monopole Antenna with Complicated Topology

395

For such structures a lot of analysis methods are well known. One group of methods is based on the determination of equivalent magnetic currents distribution around the patch edges. Most popular are: the transmission line model (TLM) [9–11]; the cavity model (CM) [12–14]; the multiport network model (MNM) [15, 16]. The methods in the second group are based on an electric current distribution on a patch conductor and a ground plane. Inherently these methods are numerical. Among them one can mention the following: the method of moments (MoM) [17, 18]; the finite-element method (FEM) [19, 20]; the spectral domain technique (SDT) [21]; the finite-difference time domain method (FDTD) [20]. A simplified approach to finding the spectrum of eigen-frequencies of a disk resonator is based on finding the roots of a dispersion equation of the form  2 (1) kρ + (kz )2 = kr2 = ωr2 μa εa , U

 represents the zeroes of the derivative of the Bessel function where kρ = amn ; Umn Jm (x), and they determine the order of the resonant frequencies; a is radius of disk; kz = ρπ d ; εa , μa are absolute dielectric and magnetic constants of dielectric substrate, correspondently. For most patch antennas, the substrate thickness d is small relative to the free space wavelength, supposing that the fields along z is constant and kz = ρπ d = 0 because p = 0. This means that the resonant frequencies for TMmn0 modes can be found by utilizing the following equation [39]     ·c Umn Umn 1 = (2) (fr )mn0 = √ , √ 2π εa μa aeff 2π aeff εr

where c is the speed of light in free space, εa = ε0 εr is absolute dielectric constant, ε0 and μ0 are dielectric and magnetic constants of vacuum, correspondently, εr is relative dielectric constant of substrate, aeff is effective radius of disk. It is given by aeff =  1+

F     0.5 , 2h πF ln + 1.7726 π εr F 2h

(3)

9

√ where F = 8.791·10 fr εr Formulas (1)–(3) allow only fairly approximately to evaluate these or those parameters of disk resonators. In addition, the presence of even the simplest inhomogeneity on the patch does not allow finding the required parameters. The parameters necessary for practical applications can be obtained only with multi-parameter numerical simulation. In this work, all the characteristics of an antenna with a complex topology will be obtained in the framework of the Finite Element Method (FEM) using the Ansoft HFSS software.

2 Design Under Study Let us consider microstrip disk resonator with one sector removed as a base structure. In Fig. 1, the structure is presented schematically. In Fig. 1 the following notations have

396

D. Mayboroda and S. Pogarsky

been used: 1 – dielectric substrate made of FLAN-3,8 with εr = 3.8 and tg δ ≈ 2 · 10−3 , substrate thickness h = 1 mm; 2 – main microstrip disk with a diameter of D = 35 mm; 3 – sector inhomogeneity with internal radius of R = 10 mm and central angle of 63.4° . The resonator is assumed to be excited by coaxial line segment. This excitation type has been chosen because removed sector makes determination of energy input point using microstrip exciter difficult. At the same time, the coaxial exciter does not impose any requirements on the energy input point position or limitations of excited oscillations spectral content.

Fig. 1. Schematically presentation of structure

As a consequence of a quite complex resonator topology, it is obviously impossible to find analytical expressions for the main antenna characteristics. Therefore, the finite element method has been chosen for numerical modeling of frequency and energy parameters. The already classical semi-open resonator model has been used [22]. The Ansoft HFSS software was used for numerical simulations. Considering the ratio between the antenna size and resonant wavelength λr , prevailing modes in this resonator can be claimed to be Emn0 (TMmn0 ). Spectral characteristic is of particular importance during studying the range properties of any functional device. Antenna devices spectral characteristics enable evaluating not only the total number of eigenmodes, but also their position on the frequency axis, which determines the operating range. It becomes especially important when the structure is axially symmetric, because in this case the modes degeneracy can arise. Spectral characteristic of disk microstrip resonator with sector inhomogeneity is shown in Fig. 2. The analysis of the characteristic shown in Fig. 2 indicates highly uneven distribution of spectral lines along the frequency axis. Within the frequency range from 0.93 to 4 GHz, the spectrum becomes quite rarefied. There are no degenerate modes in this frequency interval. Comparison with the spectral characteristic of canonical disk microstrip resonator presented in [23] indicates the eigenmodes spectrum becoming considerably sparse due to insertion of sector inhomogeneity. The influence of this inhomogeneity is less significant within the high-frequency range.

Microstrip Monopole Antenna with Complicated Topology

397

Fig. 2. Spectral characteristic

At the frequencies above 9 GHz it results in spectral lines crowding and formation of packets of spectral lines. Near the frequency 2.5 GHz there is significant spectral lines convergence, which enables prediction of elliptical (close to circular) polarization of radiated waves. At the frequencies near 4 GHz (3.96, 3.99, 4.01 GHz), degenerate modes are present, as indicated by almost complete convergence of spectral lines. 2.1 Matching Transmission line matching with radiating element has always been a quite complicated issue, especially in case of non-canonical radiating elements, elements with complex topology, etc. An input impedance of canonical disk microstrip resonator at resonance is known [24]. It depends not only on frequency parameter, but also on disk radius and energy input point coordinate: 

 0  r0  J1 1.84r a 1.84J1 a , (4) Rin = GΣ where J1 (x) − 1st order Bessel function; GΣ – conductivity of edge of resonator, defined a by GΣ = 120λ . Calculations according to these formulas enable determination Rin in 2–20 GHz range. Calculations according to previously given formulas indicate quite considerable variation of input impedance (from 120 to 240 Ohm). Even considering the average values (170…180 Ohm), matching with transmission line with the standard 50 Ohm impedance is rather difficult. In this case, the use of impedance transformer, increasing output impedance of transmission line to 170…180 Ohm, is required. The use of planar transformers turns out to be difficult in many cases, especially when conductors are placed rather densely. Besides, in matching devices, high Q-factor parasitic resonances can exist, e. g. traveling-wave resonances, which effectively absorb the incident electromagnetic wave power and prevent its radiation. Therefore, the use of coaxial line segment is preferred.

398

D. Mayboroda and S. Pogarsky

In Fig. 3, frequency dependences of |S11 | within the range 0.5–20 GHz are shown. These dependences have been simulated for different εr . values have been chosen for antennas, operating in L, S, C bands. The analysis of above dependences indicates such sectorial antenna to be multi-band. The quality of matching with transmission line turns out to be different. As εr increases, it becomes worse. It can be explained by increasing the loaded Q–factor of the microstrip resonator and consequent decreasing matching with the free space, as εr increases. The best matching corresponds to the minimum εr = 2.4. However, weak reflection from antenna input (|S11 |) does not ensure effective radiation.

Fig. 3. The dependences of |S11 | vs frequency

Fig. 4. The structure of current lines.

Figure 4 demonstrates the excitation efficiency of the radiating aperture. It shows current lines structure on the disk resonator surface, made employing color scale method.

Microstrip Monopole Antenna with Complicated Topology

399

The currents amplitude is normalized to its maximum value at the coaxial line output and is given in dB. The graph has been plotted at F = 4.85 GHz (Curve 2 corresponds to the frequency of minimum |S11 | and εr = 3.8). The oscillation mode on the sectorial inhomogeneity is rather difficult to identify, however, there are two azimuthal variations of currents amplitude on it. Furthermore, near the outer corners there are currents minima and near the inside corners there are relative maxima. The above currents lines structure predetermines the possible placement of additional inhomogeneities (e. g. slot inhomogeneities, shorting elements), which enable to change the excitation conditions and, consequently, influence both spectral and energy characteristics. 2.2 Energy Characteristics Radiation pattern and polarization characteristic are the most important operational characteristics of any radiating system. Figure 5 shows radiation patterns calculated in elevation plane of the structure with dielectric substrate εr = 2.4 at the frequencies of maximum |S11 | (Fig. 3). Curves 1, 2 correspond to F = 6.33 GHz, and curves 3,4 correspond to F = 17.19 GHz. All the dependences are normalized to general maximum. The analysis of current lines structure indicates excitation of the resonator with sectorial inhomogeneity on a slightly “distorted” E350 -mode of canonical disk microstrip resonator.

Fig. 5. Radiation pattern in elevation plane.

Curve 1 (black line) represents the radiation pattern in E-plane. The pattern has typical two lobes and a dip along the normal to antenna plane. Both lobes tilt to the normal at the angle of ±31.2◦ . The normalized level of interference lobes does not exceed 0.3 and their maximum is at the angle of 90◦ to the normal. Curve 2 (red line) represents the radiation pattern in H-plane. The radiation power in H-plane equals 0.24 of power in E-plane. Direction of maximum radiation is at the angle of ±67◦ to the normal to antenna plane. The lobes have symmetric but wrong shape. Curves 3 (green line), 4 (blue line) represent the radiation patterns in E- and H-plane respectively at the frequency F = 17.19 GHz. It should be mentioned, that the radiation power does not exceed 0.4 of maximum. The radiation pattern in E-plane is multi-lobe with two distinctive lobes and several

400

D. Mayboroda and S. Pogarsky

minor interference lobes. In H-plane the radiated power slightly increases (to 0.46 of maximum), and the radiation pattern shape becomes circular. The radiation patterns calculated in elevation plane for E-polarization for antenna with substrate εr = 3.8 at the frequencies F = 4.85 GHz and F = 13.66 GHz are shown in Fig. 6. These frequencies correspond to minimum |S11 | for εr = 3.8 (Fig. 3). Figure 7 demonstrates the radiation patterns in elevation plane for H-polarization at  the above frequencies. In each plane, E Emax has been normalized separately. These dependences can not be presented together because of significantly different radiated power in E- and H-plane.

Fig. 6. Radiation pattern in H-plane.

Fig. 7. Radiation pattern in E-plane.

In Fig. 6, curves 1, 2 have been obtained at the frequencies F = 4.85 GHz and F = 13.66 GHz respectively. The analysis of the above dependences shows that patterns have several lobes at both frequencies. Besides, as frequency increases, the number of lobes increases, whereas radiated power decreases. By comparing the radiation patterns shape at lower frequencies, it can be noticed that as εr increases, the main lobes become asymmetric and side lobes radiated power reaches 0.8 of maximum. At F = 17.19 GHz (curve 3 in Fig. 5) the radiation pattern is absolutely symmetric, whereas at F = 13.66 GHz this symmetry almost disappears and the radiated power is considerably reduced. In Fig. 7, curves 1, 2 have been obtained at F = 4.85 GHz and F = 13.66 GHz respectively. At lower frequency the radiation pattern shape is sufficiently asymmetric,

Microstrip Monopole Antenna with Complicated Topology

401

the radiation maximum is at the angle of 58.5◦ to the normal. At the frequency F = 13.66 GHz the pattern shape sufficiently changes: it becomes multi-lobe and the radiated power decreases. Polarization characteristic is another important operational characteristic. It enables to estimate the ability of antenna to radiate (or to receive) signals of different polarization. It is determined by the ellipticity coefficient (η), measured in dB. In accordance with IEEE standard, this coefficient can be calculated through the formula      |E1 |2 + |E2 |2 + E 2 + E 2  1 2   , (5) η= |E1 |2 + |E2 |2 − E 2 + E 2  1

2

where |E1 | and |E2 | are the major and the minor axes of polarization ellipse respectively. In Fig. 8, polarization dependences at different frequencies are shown. Curve 1 corresponds to F = 6.36 GHz, curve 2 – to F = 17.2 GHz.

Fig. 8. Polarization characteristic.

Obviously, in both cases there are angle ranges, within which the ellipticity coefficient does not exceed 4 dB. Furthermore, at εr = 2.4 GHz these ranges are much wider. Besides, there are bursts of ellipticity coefficient, at which the radiated wave is linearly polarized.

3 Conclusions The main frequency and energy characteristics of planar antenna based on microstrip resonator with complex topology of radiator, excited by coaxial line segment, have been studied. The high quality of operational characteristics has been achieved due to complex topology of the microstrip radiator. Excitation by coaxial line eliminates the need to find an excitation point. Besides, the excitation of almost all the resonator eigenmodes is possible. Radiated fields with specified characteristics can be formed through selection of the geometric parameters and the substrate material constants.

402

D. Mayboroda and S. Pogarsky

Acknowledgment. This work was supported by the Ministry of Education and Science of Ukraine, grants 0118U002038, 0119U002535.

References 1. Grizinov, D.B., Orlov, O.V.: Compact universal radio/television antenna”, RU Patent № 2454761 RU H01Q 1/36. Published 27.06.2012 2. Babichev, D.A., Tupik, V.A.: Fractal antenna, RU Patent № 125396 RU H01Q 13/00. Published 27.02.2013 3. Krupenin, C.V., Kolesov, V.V., Potapov, A.A., Matveev, E.N.: Multiband broadband antennas based on fractal structures of various types. Radiotechnika 3, 70–83 (2009) 4. Chen, C.H., Tulintself, A., Sorbello, R.M.: Broadband two-layer microstrip antenna. In: IEEE Trans Antennas Propagation Society International Symposium Digest, pp 251–254 (1984) 5. Griffin, J.M., Forest, J.R.: Broadband circular disc microstrip antenna. Electron. Lett. 18, 267–269 (1982) 6. Aanandan, C.K., Mohanan, P., Nair, K.G.: Broad-band gap coupled microstrip antenna. In: IEEE Trans Antennas Propagation AP-38, pp. 1581–1585 (1990) 7. Singh, K.K., Gupta, S.C.: Review and analysis of microstrip patch array. Antenna with different configurations. Int. J. Sci. Eng. Res. 4(2), 144–156 (2013) 8. Maci, S., Biffi Gentili, G., Piazzesi, P., Salvador, C.: Dual-band slot-loaded patch antenna. IEEE Proceedings Microwaves Antennas and Propagation, vol. 142, pp. 225–232 (1995) 9. MacKinchan, J.C., et al.: A wide bandwidth microstrip sub-array for array antenna application using aperture coupling. In: IEEE AP-S International Symposium Digest, pp. 878–881 (1989) 10. Menzel, W., Grabherr, W.: “Microstrip patch antenna with coplanar feed line. IEEE Microwave Guided Wave Lett. 1(11), 340–342 (1991) 11. Smith, R.L., Williams, J.T.: Coplanar waveguide feed for microstrip patch antenna. Electron. Lett. 28(25), 2272–2274 (1992) 12. Bhatacharya, A.K., Garg, R.: Generalized transmission line model for microstrip patches. In: IEE Proceedings Microwaves, Antennas Propagation, Pt. H, 132(2), 93–98 (1985) 13. Dubost, G., Beauquet, G.: Linear transmission line model analysis of a circular patch antenna. Electron. Lett. 22, 1174–1176 (1986) 14. Babu, S., Singh, I., Kumar, G.: Improved linear transmission line model for rectangular, circular and triangular microstrip antennas. In: IEEE AP-S International Symposium Digest, pp. 614–617 (1997) 15. Lo, Y.T., Solomon, D., Richards, W.F.: Theory and experiment on microstrip antennas. In: IEEE Trans Antennas Propagation, AP-27, pp. 137–145 (1979) 16. Richards, W.F., Lo, Y.T., Harrison, D.D.: An improved theory for microstrip antennas and applications. In: IEEE Trans. Antennas Propagation, AP-29, vol. 1, no. 981, pp. 38–46 (1981) 17. Lo, T., Lee, S.W.: Antenna Handbook. Van Nostrand Reinhold, New York (1988) 18. Okoshi, T., Miyoshi, T.: The planar circuit—an approach to microwave integrated circuitry. IEEE Trans. Microwave Theory Tech. 20, 245–252 (1972) 19. Gupta, K.C., Sharma, P.C.: Segmentation and desegmentation techniques for the analysis of two dimensional microstrip antennas. In: IEEE AP-S International Symposium Digest, pp. 19–22 (1981) 20. Newman, E.H., Tulyathan, P.: Analysis of microstrip antennas using method of moments. In: IEEE Trans Antennas Propagation, AP-29, pp. 47–53 (1981) 21. IE3D 7.0. Zeland Software Inc., Fremont, CA

Microstrip Monopole Antenna with Complicated Topology

403

22. Mayboroda, D.V., Pogarsky, S.A.: Electrodynamic characteristics of the disc microstrip radiator with structural auxiliary elements. Telecommun. Radio Eng. 74(13), 1147–1155 (2015) 23. Lytvynenko, L.N., Pogarsky, S.A., Mayboroda, D.V., Poznyakov, A.V.: Microstrip antenna with complex configuration of radiators. In: International Conference on Antenna Theory and Techniques (ICATT), Kyiv, Ukraine, pp. 254–256 (2017) 24. Panchenko, B.A., Nefedov, E.I.: Microstrip antennas. Radio Commun. (1989) (In Russian). 144 p

Scattering of Electromagnetic Wave By Bragg Reflector with Gyrotropic Layers Alexander Shmat’ko1(B) , Victoriya Mizernik2 , and E. Odarenko3 1 School of Radiophysics, Biomedical Electronics and Computer Systems, V.N. Karazin

Kharkiv National University, Kharkiv, Ukraine [email protected] 2 Scientific Physical-Technologic Center of MES and NAS of Ukraine, Kharkiv, Ukraine [email protected] 3 School of Electronics and Biomedical Engineering, Kharkiv National University of Radio Electronics, Kharkiv, Ukraine [email protected]

Abstract. We solved the problem of scattering of plane wave on a ferrite 1D magnetophotonic crystal controlled by a dc transverse magnetic field. Fundamental solutions of the Hill equation with mixed boundary conditions based on the Floquet-Bloch theory are obtained in an analytical form. The dispersion equation and its roots are found explicitly. The analysis of the dispersion properties of the structures is carried out depending on the material parameters of the ferrite layers. The transmission and reflection coefficients are determined for the gyrotropic crystal with finite number of periods. Two characteristic cases are considered: positive and negative values of the effective permeability of gyrotropic layer. The expressions for spatial distribution of electromagnetic field components are determined at crystal period. The results provide a deeper understanding of the electromagnetic waves propagation behavior in multilayer media with controlled gyrotropic elements. In addition, the obtained analytical expressions simplify the analysis of wave processes in such complex media. Keywords: Magnetophotonic crystal · Gyrotropic media · Hill’s equation · Floquet-Bloch theory · Ferrite with dc transversal magnetic field · Dispersion characteristics · Band gap

1 Introduction Magnetophotonic crystals (MPhC) are widely used in various applications of modern science and technology of the terahertz, microwave and optical ranges. Features of the transmission of electromagnetic waves through such structures are completely determined by the material parameters and the geometric dimensions of the layers. Dispersion properties of one-dimensional MPhC with isotropic layers are well studied, based on the obtained analytical characteristic equations for both TE and TM waves [1–8]. The most promising applications include MPhC in the presence of the gyrotropic of one or © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 M. Ilchenko et al. (Eds.): MCT 2019, LNNS 152, pp. 404–416, 2021. https://doi.org/10.1007/978-3-030-58359-0_23

Scattering of Electromagnetic Wave By Bragg Reflector with Gyrotropic Layers

405

two layers of a periodic structure. The presence of gyrotropic layer in such structure makes it possible to change the values of the material parameters of the medium relatively easily due to the applied magnetic field and, as a result, to control the dispersion properties of the periodic structure and the wave transmission coefficient electrically. For gyrotropic MPhC there are no analytical expressions for the main frequency and amplitude characteristics of the scattered fields. In the presence of gyrotropy the material parameters are tensor quantities. This is the main reason for necessity of obtaining an analytical solution for the problem of finding both the dispersion equation of a one-dimensional MPhC and the transmission and reflection coefficients of waves through an MPhC limited in the number of periods. In fact, the problem of scattering of a wave on the MPhC can be reduced to the study of the Hill equation with a discretely periodic medium. When solving the Hill equation, as a rule, the Floquet theory of waves [9–12] is used. Such problems were considered only for the case of isotropic two-layer discretely periodic media [13–23] where analytical expressions for the Floquet-Bloch functions in the layers were obtained. For gyrotropic MPhC, the Floquet theory approach was not considered. In this paper, on the basis of the Floquet theory and the Hill equation, the problem of the scattering of plane wave on a bounded ferrite crystal with a controlling transverse magnetic field is solved analytically.

2 Floquet - Bloch Waves Currently, MPhC are widely used in various devices for transmitting information in the terahertz range. To calculate the amplitude-frequency characteristics of such structures, the transfer matrix method [9–12, 14] is used usually. The most general case of calculating the dispersion characteristics in MPhC with bigirotropic layers using the transfer matrix method is given in [3, 6]. In this report, the problem of scattering of the E-polarization plane wave on a limited MPhC consisting of a finite number of periods of the ferrite and dielectric layers is considered. Scheme of structure under investigation and appropriate coordinate system are shown on Fig. 1. The problem is reduced to the Hill equation with variable coefficients, the solution of which is based on the Floquet theory. Let us turn to the solution of the problem. We will consider the scattering of TM waves with a component of the magnetic field Hz = 0 (Hx , Hy , Ez ), Ez -polarization (s-polarization). Maxwell equations are reduced to the Helmholtz equation for the electric field component:   1 ∂Ez 1 ∂ 2 Ez ∂ + + k 2 εj (x)Ez = 0 (1) ∂x μ⊥ (x) ∂x μ⊥ (x) ∂y2 where μ⊥ (x) and ε(x) are determined by the known formulas for the gyromagnetic ferrite layer and the dielectric layer [13], namely:    μ −iμ 0  j aj   ↔   0 , μj =  iμaj μj    0 0 μj 

406

A. Shmat’ko et al.

Fig. 1. Schematic of the ferrite Bragg structure.

  where μ⊥j (x) = μj 1 − μ2aj /μ2j is effective values of magnetic permeability of ferrite layers, εj (x) = εj is dielectric constant of layers. Relationship between the field components Hy , Ez is determined by the expression:    ∂Ez 1 μa ∂Ez . (2) Hy = − +i ikμ⊥ ∂x μ ∂y It should be noted that the principle of permutation duality for the two types of waves (TM and TE) follows from Maxwell equations. Therefore it suffices to consider a single polarization of the waves. Solution for another polarization field (e.g. for Hz ) is found by ↔ ↔ simply replacing the material parameters μ ↔ − ε . All this allows us to simplify the consideration of a general electrodynamics problem and restrict ourselves to only one type of TE or TM wave for any type of medium. Equation (1) can be reduced by using the method of separation of variables Ez (x, y) = X (x)eβy to one kind of Hill equations with periodic coefficients, namely:   ∂X ∂ p(x) + q(x)X = 0, (3) ∂x ∂x where p(x) and q(x) are periodic coefficients determined by expressions: 1 , μ⊥j (x)   q(x) = p(x) k 2 εj (x)μ⊥j (x) − β2 = p(x)ξ2 (x) = pj ξ2j p(x) =

Scattering of Electromagnetic Wave By Bragg Reflector with Gyrotropic Layers

407

for TM waves. The value β determines the wave propagation constant along the axis Oy(exp(±iβy)) and associated with the operator ∂ 2 /∂y2 = −β2 in Helmholtz Eq. (1). We assume that the layout of the layers is as follows: a layer with an index j = 1 occupies areas of space nL < x < a + nL, layer with index j = 2 occupies additional area to the structure period a + nL < x < L + nL. Here n = 0, 1, 2 . . . is period number. The one-dimensional Eq. (3) is the Hill equation [9–12, 15–23] with periodic functions p(x) and q(x), such that p(x + L) = p(x),q(x + L) = q(x), where L is MPhC period. Equation (3) with the corresponding boundary conditions is the boundary SturmLiouville problem. According to Hill equation theory, if ψ1 (x) is a particular solution of Eq. (3) with periodic coefficients then ψ1 (x + L) is also solution. If ψ1 (x) and ψ2 (x) are two linearly independent solutions of Eq. (3) then ψ1 (x + L) and ψ2 (x + L) are also a solution of Hill equation. These solutions can be represented as a linear combination of two fundamental solutions ψ1 (x) and ψ2 (x), namely:  ψ1 (x + L) = a11 ψ1 (x) + a12 ψ2 (x) , (4) ψ2 (x + L) = a21 ψ1 (x) + a22 ψ2 (x) where anm are constants to be defined. Therefore shift on the period L along coordinate axis Ox is reduced to linear combination of fundamental solutions ψ1 (x) and ψ2 (x). Two linearly independent fundamental solutions of Hill equation are chosen in such a way as to satisfy the boundary conditions: ψ1 (0) = 1, ψ1 (0) = 0, ψ2 (0) = 0, ψ2 (0) = 1. Using these boundary conditions for ψ1 (x) and ψ2 (x) constants in (4) can be obtained as follows: a11 = ψ1 (L), a21 = ψ2 (L), a12 = ψ1 (L), a22 = ψ2 (L). Then the characteristic equation for determining a Floquet constant ρ takes the standard form: ρ 2 − 2Aρ + Wr(ψ1 , ψ2 ) = 0,

(5)

where A=

1 ψ1 (L) + ψ2 (L) = cos ϕ, 2

Wr(ψ1 , ψ2 ) = ψ1 (x)ψ2 (x) − ψ2 (x)ψ1 (x). Naturally, in its meaning, the roots of the characteristic equation should not depend on the choice of fundamental solutions ψ1 (x) and ψ2 (x). Free term Wr(ψ1 , ψ2 ) in (5) is constant because it determines the Wronskian of the original equation. The introduction of the above boundary conditions to determine the fundamental solutions is connected with the well-known Dirichle and Neumann boundary value problems [15–23]. With regard to the MPhC with gyrotropic layers, such boundary conditions are not optimal from the point of view of the simplicity of finding the eigenfunctions of the Hill equation. In this case, to find fundamental solutions it is advisable to use mixed boundary conditions. This follows from the form of the field (2). Boundary conditions for finding

408

A. Shmat’ko et al.

solutions of Hill equation X (x) = Aψ1 (x) + Bψ2 (x) connected to the continuity of the tangential field components on the boundaries of layers:    ∂X (x) 1 μa Ez (x, y) = X (x)eiβy , Hy (x, y) = − β X (x) , −ikμ⊥ ∂x μ and can be written as follows: X1 (a) = X2 (a),     1 ∂X1 (a) ∂X2 (a) μa1 μa2 1 −β −β X1 (a) = X2 (a) . μ⊥1 ∂x μ1 μ⊥2 ∂x μ2

(6)

The characteristic equation for a ρ value can be obtained by using the Floquet theorem on the MPhC period, namely: ρX1 (0) = X2 (0 + L),     ∂X1 (0) ∂X2 (0 + L) 1 1 μa1 μa2 ρ X1 (0) = X2 (0 + L) . −β −β μ⊥1 ∂x μ1 μ⊥2 ∂x μ2

(7)

We will find fundamental solutions of the Hill equation as a solution of third boundary value problems with mixed boundary conditions 

μa1 1 ∂ψ1 (0) ψ1 (0) = 1, −β ψ1 (0) = 0, μ⊥1 ∂x μ1   ∂ψ2 (0) μa1 1 ψ2 (0) = 0, −β ψ2 (0) = 1. μ⊥1 ∂x μ1 The fundamental solutions of the Hill equation, taking into account the mixed boundary conditions, take the following explicit form for two regions on the period of a onedimensional MPhC: cos ξ1 x + β μμa11 sinξξ1 1 x 0