A Semi-Supervised Clustering Method For P2P Traffic Classification

In the last years, the use of P2P applications has increased significantly and currently they represent a significant portion of the Internet traffic. In consequence of this growth, P2P traffic identification and classification are becoming increasingly important for network administrators and designers. However, this classification was not simple. Nowadays, P2P applications explicitly tried to camouflage the original traffic in an attempt to go undetected. This paper present a methodology and selection of three P2P traffic metrics and applies semi-supervised clustering to identify P2P applications. Three P2P traffic metrics: IP Address Discreteness, Success Rate of Connections and Bidirectional Connections rate had been proposed and used in this paper. The semi-supervised classification method for P2P traffic consist two steps: Particle Swarm Optimization (PSO) clustering algorithm was employed to partition a training dataset that mixed few labeled samples with abundant unlabeled samples. Then, available labeled samples were used to map the clusters to the application classes. Experimental results using traffic from campus showed that high P2P traffic classification accuracy had been achieved with a few labeled samples.