CASPIAN JOURNAL
MANAGEMENT AND HIGH TECHNOLOGIES
Neural network technology for detection of network intrusions
 |
Read | Kochetov D. A., Lukashchik E. P. Neural network technology for detection of network intrusions // Caspian journal : management and high technologies. — 2018. — №2. — pp. 104-112. |
Kochetov D. A. - undergraduate student, Kuban State University, 149, Stavropolskaya St., Krasnodar, 350040, Russian Federation, aziris_nuna@list.ru
Lukashchik E. P. - Cand. Sci. (Physics and Mathematics), Associate Professor, Kuban State University, 149, Stavropolskaya St., Krasnodar, 350040, Russian Federation, lep_9091@mail.ru
The efficiency of an information protection system in a computer network depends on the type of technology chosen for the detection of network intrusions. At present, the neural network technology is the most promising for these purposes. Having been trained, synthetic neural networks are able to adapt to new types of threats and recognize them even if they did not detect them before. This feature allows a protection system based on synthetic neural networks to become more flexible and independent. This article demonstrates the application of the neural network technology for building a protection system. This technology can capture a data stream from the network, analyze these data and, in case of a threat, inform the administrator. A number of simulation experiments were carried out, including the creation of different types of neural networks, its training with the use of different methods and testing for security efficiency assessment. As a result, the neural network architecture that classifies input data most accurately and fully has been chosen. A virtual server was deployed at the hosting provider Flops to test the software of the developed detection system. The traffic from different services was directed to that server. The results of computer-aided experiments for different types of attacks prove the efficiency of neural network method to solve the problems of safety of the computer information system based on a computer network.
Key words: информационная безопасность, сетевая атака, обнаружение вторжений, нейронные сети, сети Кохонена, многослойный персептрон, Python, вычислительные эксперименты, information security, network attack, intrusion detection, neural networks, Kohonen networks, m