Using Traffic Self-Similarity for Network Anomalies Detection

The goal of this paper is to prove the potential of fractal analysis techniques in evaluation of network characteristics, especially in detection of anomalies, as a method to reveal self-similarities in generated traffic. After a short review of some anomaly detection methods, one describe in detail a statistical signal processing technique based on abrupt change detection. A case study based on real network data from the database of management variables of a SNMP server demonstrates the power of the signal processing approach to network anomaly detection.