A novel statistical automaton for network cloud traffic classification

Traffic classification is crucial in many network and cloud applications, they are from QoS enforcement, network monitoring to security and firewalls. In recent years, all the classification with deep packet inspection (DPI) are using the exact matching with the existing policy semantics. However, if the policy semantics is changed, then the DPI classifier is no longer able to be a workable traffic classification. We proposed a new statistical automaton for the traffic classification. The applications are marked by many multiple signatures during a flow training process, and then it classifies the applications when their statistical results are reached. In the experiment, we evaluate the proposed method with 5 applications which proves our idea is feasible for the network and cloud traffic classification.