Anomaly Detection and Localization by Diffusion Wavelet-based Analysis on Traffic Matrix

Teng Sun, Hui Tian, Xuan Mei

Diffusion wavelets (DW) transform has been successfully used in Multi-Resolution Analysis (MRA) of traffic matrices because it inherently adapts to the structure of the underlying network. There are many potential applications based on DW analysis such as anomaly detection, routing optimization and capacity plan, which, however, have not been well developed. This paper shows how to apply two-dimensional DW transform in traffic matrix analysis and anomaly detection. The experimental results demonstrate the effectiveness of DW-based technique in traffic matrix analysis and anomaly detection in practical networks. It also shows this new technique is potential to be used in many other network applications.