Research And Implementation Of P2P Traffic Detection And Control System

In recent years, the popularity of P2P technology, rich Internet applications, especially file resource sharing and distributed computing based on P2P network fully display its advantages. But it also brings the problems:on the one hand, P2P applications take up a lot of bandwidth resources caused by network congestion; on the other hand, the complexity of the P2P network to bring the issue of network security is becoming increasingly prominent. Therefore, it is especially important to accurate identification the P2P traffic for solving these problems.The paper first analyzes the P2P technology, as well as P2P traffic identification technology, and summarizes the advantages and disadvantages of the current P2P traffic identification method. And then determine based on deep packet inspection technology and P2P traffic identification method based on machine learning as the focus. Based on deep packet detection technology research, first capture and reverse analysis method to obtain the characteristics of commonly used P2P protocol application layer, then the analysis of a typical string matching algorithm, select the matching algorithm more practical, finally, the identification of common P2P protocol. P2P traffic identification based on machine learning research, first feature selection method based on mutual information filter out the classification feature subset from248kinds of features in Andrew Moore’s datasets, the subset excludes redundant features. Then research the random forest classification algorithm, which is built using bagging method. Finally, the feature subset is used by the random forest classifier, the experimental results show that:the selected features based on mutual information method for RF classifier achieved higher flow recognition accuracy.On the basis of the analysis and research, research P2P traffic detection and control model for the campus network. First studied the Netfilter architecture of the Linux system, and extend the Netfilter framework for base of the system. Then divide the system module and use the thought of system, and that based on deep packet inspection technology and machine learning-based identification method will be used to P2P traffic identification module. Complete the detailed design of each module. Finally implement the system, and analysis and evaluation the system by experiment. The experimental results demonstrate the effectiveness of the system.