Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Computer network > General issues > Computer networks, test , run

DFI Research in P2P Traffic Identification

Author LiuTao
Tutor ChenHongWei
School Hubei University of Technology
Course Applied Computer Technology
Keywords P2P DFI Bayesian Decision tree BP Neural NetWork Flow characteristic
CLC TP393.06
Type Master's thesis
Year 2010
Downloads 60
Quotes 0
Download Dissertation

With the Internet rapid development, P2P applications in the network is becoming the main occupants of broadband networks and has also become an important component of network services. P2P technology in promoting the development of the Internet network, but also occupy on network bandwidth resources, network congestion, network copyright, network traffic, occupancy, and lead to an increase the difficulty of network management, security risks and other issues, these problems will hinder the normal network operations.we could effectively identify P2P applications on the context of management control, P2P application detection technology is mainly deep packet inspection and depth of the flow assay. Deep packet inspection method can not detect encrypted and unknown application of peer network flowing, the depth of flow detection method can overcome the deep packet inspection method of this shortcoming. In this paper, based on the depth of deep packet inspection identify P2P traffic detection flow applications. This paper describes the P2P technology, the basic theory discussed P2P technology works, technical features, and detailed traffic detection technology in depth and a fuzzy algorithm for the three algorithms applied research.In the above, based on research and analysis, this paper presents an application to identify unknown P2P applications research skills. The system uses fuzzy Bayesian algorithm, decision tree, neural network, fuzzy detection. Three kinds of algorithms are adopted out of the results of deep packet inspection as an ideal output set, using the same characteristics of network traffic, only the characteristics of the categories contain enough information to allow the use of classifiers to achieve the correct classification. In the extracted features, there will be three kinds of algorithms, and concluding with the analysis results is displayed.The two pairs of Bayesian classifiers-Naive Bayesian and Full Bayesian algorithm, the training results, run the results of research, experimental studies have shown that naive Bayes and full Bayesian classifier can be quickly and accurately find P2P streaming applications, Naive Bayesian classifier accuracy dominate the whole time to take advantage of Bayesian run. For the decision tree, decision tree classification techniques experiments show that more rapid and accurate positioning of P2P data stream, in real-time processing large amounts of data in decision tree classification techniques to better reflect the efficiency and accuracy. For the neural network, experiments show that neural network to detect the accuracy of P2P data stream is higher, in real-time processing large amounts of data in some less efficient.Test results show that the system can effectively detect known or unknown network P2P application traffic, contribute to the network bandwidth resource management, ensure the normal operation of conventional operations.At the end of the paper, it summarized the work and prospected for the work direction of next step.

Related Dissertations
More Dissertations