Using Packet Size Distribution to Identify P2P-TV Traffic

Abstract

We propose an approach to accurately identify the traffic generaed by P2P-TV applications. This approach only relies on accounting packet size distribution (PSD) of the P2P-TV application during small time-windows. The rationale is that the packet size distributions of P2P-TV applications are different from each other. Based on the statistic, we make use of Support Vector Machines on a large set of P2P-TV test-bed traces. Experiment results show that our approach is effective and reliable to identify P2P-TV traffic.