Research On The Method Of P2P Network Traffic Classiifcation Based On Behavioral Characteristics

Nowadays P2P technology has made great changes on the Internet, it changes thetraditional Server-Client model of network, and gives people a brand new way toacquire resources, it has been widely applied. However, P2P applications have broughtthe rapid growth of network traffic, resulted in a serious consumption of the networkbandwidth, and caused a great impact on the quality of service for other network traffic.Therefore, how to accurately make classification of P2P traffic is becoming a hot issuein current research.Basic principles and methods of machine learning are studied in this thesis, andprinciple of Support Vector Machines and classification method are focuses on, methodsfor network traffic and P2P traffic classification are analyzed. The behavioralcharacteristics of the P2P network application, behavior-based attributes and P2Pnetwork traffic classification model based on support vector machines are given. Themodel is simulated on LIBSVM tools with sample data of P2P application traffic. Themain works of the thesis are as follows:(1) The basic principles and methods of machine learning are studied, the principleof Support Vector Machines and classification methods are discussed. The naiveBayesian classification, decision tree classification, clustering classification and neuralnetwork classification methods of port-based P2P traffic classification are analyzed andDeep Packet Inspection classification method for P2P traffic based on the behavioralcharacteristics are researched.(2) P2P network flow behavior characteristics, P2P network traffic classificationmodel based behavioral characteristics and Support Vector Machine are given. Thebehavioral characteristics of several common P2P applications are analyzed. Thefunctional characteristic, data traffic characteristic and packet characteristic for P2Papplications are defined. The behavioral characteristics inspection model and it’sinspection algorithm are given.(3) A P2P network traffic classification model based on traffic behavioralcharacteristics and SVM classification mothed is brought out. The SVM kernelfunctions and parameters are given, and the P2P traffic classification model based onSVM model are designed.(4) On the basis of LIBSVM tools and sample data for P2P traffic, the classificationmodel is simulated, and its performance are discussed and analyzed.