Real network traffic anomaly detection based on Analytical Discrete Wavelet Transform

Signal processing techniques have attracted a lot of attention recently in the networking security technology, because of their capability of detecting novel intrusions or attacks. In this paper, we propose a new detection mechanism of network traffic anomaly based on Analytical Discrete Wavelet Transform (ADWT) and high-order statistical analysis. In order to describe the network traffic information, we use a set of features based on different metrics. We evaluate our technique with real traffic dataset, collected over several days period on a university public server. The test results show that the proposed approach accurately detects a wide range of anomalies.