Network Traffic Anomaly Detection Based on Self-Similarity Using HHT and Wavelet Transform

Abstract

Network traffic anomaly detection can be done through the self-similar analysis of network traffic. In this case, the abnormal condition of network can be indicated by investigating if the performance parameters of real time data locate at the acceptable ranges. A common method of estimating self-similar parameter is the Wavelet transform. However, the Wavelet transform fails to exclude the influence of non-stationary signal’s periodicity and trend term. In view of the fact that Hilbert-Huang Transform (HHT) has unique advantage on non-stationary signal treatment, in this paper, a refined self-similar parameter estimation algorithm is designed through the combination of wavelet analysis and Hilbert-Huang Transform and a set of experiments are run to verify the improvement in the accuracy of parameter estimation and network traffic anomaly detection.