An HTTP Video Stream Identification Method Based on Wavelet Packet Analysis and SVM

Refined traffic identification is the basis for operators to reasonably manage the network while effective feature extraction is a crucial part of it. As the dominant component of traffic, video traffic should be studied carefully. In this work, an HTTP video stream identification method making innovations in feature extraction utilizing wavelet packet analysis is proposed, and the Support Vector Machine(SVM) is applied. By regarding traffic as a signal, some frequency-domain features can be extracted. As an experimental analysis, our proposed method is validated from a publicly available dataset of real network traffic. The qualitative and quantitative experimental results show that the classifier with the union of time-domain features and frequency-domain features outperforms the one with time-domain feature set in the conventional methods in terms of accuracy, F-measure, precision and recall.