Research And Implement On P2P Streaming Identification Method Based On Behavior Analysis

With the continuous development of P2P technology, the security problem of P2P network is increasingly apparent, also, the network’s characteristics of abuse of bandwidth is taken more and more attention by the telecom operators. Therefore, how to classify and identify traffic generated by P2P services has become an urgent problem and meaningful topic.Although there are already some results on the P2P file download service recognition research, but in the field of P2P streaming identification, there is no more effective recognition algorithms, and the research on the behavior of P2P streaming traffic characteristics is very few. In real networks, especially in recent years, the growth rate of P2P streaming traffic is much higher than other types of P2P applications. The protocol encryption is an inevitable trend. Therefore, it is important to design and implement the P2P streaming traffic identification system, which based on research on P2P streaming traffic and study of behavioral characteristics.Due to its own features of architecture, P2P streaming applications have demonstrated some obvious characteristics which distinguish them from other applications in the flow. Currently, the main structure of majority P2P streaming systems is based on the mesh-pull system architecture, which determines the nodes in the system will have their own unique characteristics while they interact with others; In addition, although there are many kinds of P2P streaming applications, but no matter what kind of application, its most essential feature is the feature of temporary resource and dynamic structure. As a result, the paper selected these features as basis of traffic identification. We start a research based on the selected features of P2P streaming traffic, and design an appropriate recognition algorithm of P2P streaming traffic recognition.In this paper, the structure of P2P streaming system is described first, and the interaction behavior characteristics of nodes in P2P streaming system are summarized; Then we design and implement two identification algorithms:identification algorithm based on Bayesian classification and identification algorithm based on features matching, both of them are proposed according to the analysis on behavior model; Finally, we set up an experimental platform to verify the recognition algorithms, in this platform, two algorithms are combined as an integrated algorithm. We choose three P2P streaming applications which have most popularity in China, in order to generate traffic as the sample for the system. A large scale of experiments is given, and the results are listed. The test results show that the algorithm can effectively identify P2P streaming traffic in the unknown input traffic, and the algorithm perform excellent on recognition rate, false recognition rate, and real-time feature.