Proposion and Analysis of a TCP Feature of P2P Traffic

Over the last years, the wide use of the P2P application has lead to the rapid growth of network traffic. Thus, the accurate classification of P2P traffic becomes a challenging problem. This paper proposes some new fundamental characteristics of TCP traffic to achieve its discrimination. To prove the universality of the values of our proposed features, we also present some analytic estimates of them using Pareto as the distribution of user lifetime in real P2P systems. Finally, we apply SVM to the discrimination of these features in order to automate the processing of the discrimination, and its accuracy is demonstrated as a result of some experiments.