Multi-levels traffic classification technique

Classifying Internet network traffic flows offers substantial benefits to a number of key areas in network engineering and surveillance. However, as many newly-emerged P2P applications use dynamic port numbers and masquerading techniques, it causes the most challenging problem in network traffic classification. In this paper, we propose a novel multi-levels traffic classification technique that brings together the benefits of port mapping, signature matching and flow statistical classification techniques, motivated by variety of network activities and their requirements of traffic. Experiment results illustrate this technique can achieve high accuracy, low overheads, robustness, and real-time.