Anomaly recognition in bursty IP traffic models

In the first part of this article we deal with two models of IP traffic with bursty period, Poisson process and Pareto process. The paper deals with analysis of their probability structure and their comparison with each other. In the second part we deal with the use of one-parameter machine learning methods (autoregression coefficient and Hurst coefficient) for the recognition of anomalies in the traffic flow. We have created several scenarios for this task. These scenarios model changes of the observed flow structure. This article presents implementation of the Department of InfoComm Networks research, which focuses on computer security, into the curriculum of the master's study program Applied Network Engineering.