Intrusion Detection of Traffic Realizations Based on Maching Learning Using Fractal Properties

In this article we consider the problem of intrusion detection in computer networks by the realizations of network traffic. To solve this problem, time series analysis methods, fractal methods and data mining are used. The verification of anomaly detection methods was performed using experimental data sets. Using these data sets, realizations of multifractal stochastic binomial cascades were generated with a change of the multifractal characteristics. In the work, numerical experiments were conducted to detect intrusions using the example of detecting typical DDoS attacks, which showed that the Random Forest method using regression decision trees can effectively detect anomalies in network traffic.