A New Method of P2P Traffic Identification Based on Support Vector Machine at the Host Level

These years, P2P applications have multiplied, evolved and take a big part of Internet traffic workload. Identifying the P2P traffic and understanding their behavior is an important field. Some port, payload and transport layer feature based methods were proposed. P2P traffic identification methods by examining user payload or well-defined port numbers no longer adapt to current P2P applications. In recent years, some scholars do researches on traffic classification by using Machine Learning. However, previous researches are almost on the flow level. In this paper, we develop a new method of P2P traffic identification based on Support Vector Machine by analyzing packet length, remote hostspsila discreteness, connection responded success rate and the ratio of IP and port at the host level without relying on the port numbers and packet payload. Finally, the experiment results indicate that this approach can effectively identify P2P applications.