Problems of training deep neural networks to detect security threats in networks with dynamic topology

At present, the introduction of computer networks with dynamic topology is becoming a ubiquitous phenomenon. In everyday life, we often encounter them without knowing it. Mobile, road, sea and air dynamic networks are everywhere, and their distinctive feature is the constant change in the structure due to the constant updating of the end nodes in the network. Due to such a wide spread in these networks, there are a sufficient number of security threats both at the hardware level and at the software level. Such threats cannot be ignored. In this regard, this paper is devoted to the consideration of the main threats of security breaches at the software and network levels in networks with dynamic topology and the problems that arise when training a deep neural network to detect these threats. The analysis of the problems of training deep neural networks is carried out and the method of their elimination is proposed using the studied methods of solving such problems. As a result of the practical implementation of the technique, it is possible to obtain a properly trained neural network that will effectively detect security threats in real time.

1. Information security. Basic terms and definitions: GOST R 50922-2006, instead of GOST R 50922-96. 2008:1-5. Available at: http://www.consultant.ru. (accessed 15.12.2020). (In Russ)

2. Demidov R. A. Identification of threats to information security violations in networks with dynamic topology using deep learning methods. Dissertation for the degree of Candidate of Technical Sciences. 2018:1-143. (In Russ)

3. Neural network. Online modeling. Available at: http://primat.org/demo/network/network.html#1. (accessed 17.12.2020). (In Russ)

4. Neural networks and deep learning, Chapter 1: Using neural networks to recognize handwritten numbers. Available at: https://habr.com/ru/post/456738/. (accessed 10.12.2020). (In Russ)

5. Basic model of threats to the security of personal data during their processing in personal data information systems (extract): approved by the Deputy Director of the FSTEC of Russia 15.02.2008. 2008: 1-8. Available at: http://fstec.ru/component/attachments/download/289. (accessed 10.12.2020). (In Russ)

6. Vorobyev L. V. Systems and networks of information transmission: a textbook for students of higher educational institutions. 2009:1-336. (In Russ)

7. Goldstein B. S. Communication networks: a textbook for students of higher educational institutions. 2010:1-400. (In Russ)

8. Information security. Ensuring the security of telecommunications networks. General provisions: GOST R 52488-2005. 2007: 1-7. Available at: http://www.consultant.ru. (accessed 13.12.2020). (In Russ)

9. Information security. Vulnerabilities of information systems. Classification of information system vulnerabilities: GOST R 56546-2015. 2016:1-17. Available at: http://www.consultant.ru. (accessed 13.12.2020). (In Russ)

10. Information technology. Methods and means of ensuring security. Criteria for assessing the security of information technologies. Part 1. Introduction and general model: GOST R ISO IEC 15408-1-2012 instead of GOST R ISO IEC 15408-2008. 2013:1-56. Available at: http://www.consultant.ru. (accessed 14.12.2020). (In Russ)

11. Krukhmalev V. V., Gordienko V. N. Fundamentals of building telecommunications systems and networks: a textbook for students of higher educational institutions. 2004: 1-510. (In Russ)

12. Sokolov A.V. Information protection in distributed corporate networks and systems. 2002:1-656. (In Russ)