Weight Based Multiple Support Vector Machine Identification of Peer-to-Peer Traffic

Abstract

These years, P2P applications are very popular on the Internet and take a big part of the Internet traffic workload. Identifying the P2P traffic and understanding their behavior is an important field. Previous P2P traffic identification methods by examining user payload or well-defined port numbers no longer adapt to current P2P applications. In this paper, we develop a Multi-SVM based P2P traffic identification approach by analyzing the data transmission mechanism and connection characteristics of P2P networks at the transport layer without relying on the port number and packet payload. The result shows that the approach proposed in this paper can identify P2P traffic accurately.