ABSTRACT
Since the past decade, peer-to-peer (P2P) applications are popular among the users and they continue to be one of the largest contributors to the Internet traffic. Such applications generate a lot of traffic in the network and pose various issues to the Internet service providers (ISPs) and network administrators such as dealing with network congestion, security, maintaining quality of service for various applications, etc. This raises a need to monitor and classify the Internet traffic generated by P2P applications. Therefore, this field is actively researched as new application protocols keep on emerging. Traditional classification techniques such as port based and payload based are inefficient in classifying such traffic due to their various limitations. File-sharing traffic is the largest contributor of the P2P Internet traffic. Hence, this paper focuses on classifying P2P file-sharing traffic by utilizing a combination of heuristic based and statistical based techniques. We identified a set of heuristics and unique packet size distribution of P2P file-sharing traffic using real offline traffic datasets. The experimental results show that the proposed technique is able to achieve classification accuracy over 98.5%. In addition to that, it works on both TCP and UDP protocols, uses minimum heuristics for classification (and hence has less overhead) and can be used for real-time classification as well.
