Most Wanted Internet Applications: A Framework for P2P Identification

Abstract

For almost a decade, Peer-to-Peer (P2P) traffic have been putting pressure on network operators. Taking actions to control P2P traffic is a daunting task. In this paper, we present a comprehensive framework for the identification of P2P traffic based on information-theoretic techniques. Despite the inherent difficulty to single out such applications, our methodology is able to successfully identify P2P traffic using a set of communication patterns or profiles. We show that profiles built on the observation of traffic volume are more accurate than those using the number of flows.