MultiLevel P2P Traffic Classification Based On Behavior Feature

Nowadays P2P technology has caused a great change on Internet,it alters the server-centered pattern of network,makes a large number of Internet users participate in networks sharing and explores the potential of online resources efficiently,it gives people a brand new way to acquire resources.However,alone with the increasingly growing P2P traffic,it has become a problem to manage the traffic.Networks are overburdened ,which affects the quality of other online services.In view of the fact that the development of P2P is unstoppable,ISPs have to deal with this problem.It's getting more and more important for us to find ways to classify P2P traffic.P2P applications don't have a universal protocol,and because of their complicated developmental cause,many current P2P softwares apply all kinds of technology to avoid being detected,which makes P2P traffic difficult to be classified.Traditional methods such as port scans or DPI would probably fail in this case.This thesis makes a research of several popular P2P applications,analyses various phases of differences between traditional and P2P applications,and summarizes the commonness of P2P traffic,then infers behavior features from connection patterns,package characteristics,etc,and presents a multilevel method based on behavior feature to classify P2P traffic in a way combining traditional methods and machine learning. This method needs less information to reach a conclusion,and has ability to classify the traffic encrypted.There are seven chapters in this paper,Chapter 1 provides a brief introduction of the research background and the main content of the paper. Chapter 2 introduces the history of P2P and several methods which are used to classify P2P traffic.Chapter 3 gives a analysis of the most widely used P2P protocol BitTorrent. Chapter 4 provides analysis of different aspects of P2P behavior feature.Chapter 5 presents a method to achieve P2P traffic classification.Chapter 6 is the test report of the classification method this paper presents.Chapter 7 is the summarization of the research work and points out the next step to expand the study.