IDENTIFICATION AND PREDICTION OF SELF-SIMILAR TRAFFIC OF INFORMATION AND COMMUNICATION NETWORKS FOR ATTACK DETECTION SYSTEMS

Abstract

The urgency of development and improvement of intrusion detection systems is determined in the work, the main task of which is to recognize network attacks, attempts of unauthorized access and use of network resources. This problem is solved by using monitoring tools that can analyze real-time network traffic. For this purpose the mathematical problem of structural-parametric identification and forecasting of traffic in information and communication networks is formulated. The efficiency of solving the problem of structural-parametric identification of traffic using global optimization methods and intelligent basic functions is shown by modeling on the basis of experimental data. The efficiency of using block-oriented structures of the required models and adaptive fuzzy algorithms in solving the problem of traffic forecasting is proved. The adequacy of the obtained models of network traffic with experimental data is checked and confirmed. The aim of the work is to study and substantiate predictive models of network self-similar traffic to detect its anomalies when used in systems to detect and prevent attacks. The methodology of solving this problem is the integrated use of methods of artificial intelligence systems (neural networks, fuzzy inference systems, fuzzy clustering), structural (evolutionary, exploratory) and parametric (gradient, quasi-Newtonian) optimization methods, statistical methods for modeling and experimental data. modern traffic of information and communication networks, which has a self-similar nature. Scientific novelty. The method of structural-parametric identification and forecasting of self-similar traffic of information and communication networks is substantiated, which includes the composition of global and local optimization methods, as well as the choice of block-oriented model structures, which allows to increase the probability of traffic. Conclusions. The mathematical problem of structural-parametric identification and forecasting of traffic in information and communication networks is formulated.