Research On P2P Streaming Media Traffic Identification Techniques Based On Application Signatures

With the fast development of P2P networking and multimedia information processing technology, P2P streaming media application achieved by the adoption of P2P technology is gradually becoming another popular application favored by Internet users following P2P file-sharing application. Also the hidden safety troubles are severely emerging along with the rapid development of P2P streaming media service, so it is necessary to monitor P2P streaming media effectively and legally. P2P streaming media traffic identification is the foundation of P2P streaming media traffic monitor. Therefore, the study of P2P streaming media traffic identification has been becoming a key issue.As the study of P2P streaming media traffic identification is still in its infancy and all the related studies focus on the P2P traffic identification, so the study of P2P streaming media traffic identification faces many challenges. In this thesis, the author goes deep study on P2P streaming media traffic identification and brings up effective identification methods based on the study of P2P traffic identification technology. More detailedly, it makes the following contributions:Viewing the simplicity and accuracy of application signatures for P2P traffic identification, and also its capability of application-level classification, this thesis proposes an effective P2P streaming media traffic identification method based on applicaton signature features. These features are extracted by us from five mostly popular P2P streaming media platforms, such as PPLive, SopCast etc. Experimental results prove these features to be effective for P2P streaming media traffic identification through an application signatures based approach proposed in this thesis.Since P2P streaming media always uses private protocol, it is difficult to extract signatures manually. And feature comparison through string-matching is very expensive. Moreover, the signatures may change along with the time. All these show that automatic signature extraction methods are more preferable. Thus the thesis presents a BP neural network based approach, which can automatically select signature features for different P2P streaming media platforms via training process and meet the needs of real-time applications. It is also capable of application-level classification of P2P streaming media traffic. Both binary and multiple classification are studied. Experimental results verify its effectiveness and efficiency.Lastly, the thesis applies the methods in thesis to design a P2P streaming media traffic identification system based on the modular design principles.P2P streaming media traffic identification is the core part of P2P streaming media traffic monitor. So the study in thesis has laid good theory and technology foundation for P2P streaming media traffic monitor.