Research Of Indentifying The Early Stage Of P2P Traffic Base On SVM

P2P applications improve the quality of service,mean while they have brought new challenges for the management of network traffic.The identification of P2 P traffic is widely concerned by researchers.In this paper,the identification by early sta ge of P2 P traffic based SVM is researched.The early stage of network flow contains a wealth of identification information.The recognition of early stage of traffic is very important for the improvement of traffic identification.SVM is an excellent machine learning algorithm with excellent learning and generalization ability.It is very suitable for identification of applications traffic.In the paper,the research is carried out for the identification of early stage P2 P traffic based on SVM.The research contents of this paper include:1.By Research on P2 P Application.A framework based on SVM for identification of early stage P2 P traffic is proposed.And through the experiment,several common P2 P application traffic to identify,verify the feasibility and effectiveness of the framework.2 The process of network data preprocessing in the framework is designed.Through the collection,data purification,network flow conversion,feature statistics and other steps to transform the data from packets to samples.And the process of learning machine generalization model is optimized.3.Research on the feature of early stage of P2 P traffic.Argument the relationship between it and the feature selection.This paper compares the three feature selection strategies based on packet,flow,data packet-flow hybrid.It is proved that the identification by early stage P2 P of traffic based on packet-network flow hybrid feature is superior to based on packet or flow in comprehensive performance of accuracy,real-time and stability.More suitable for practical problems.