摘要： 针对点对点(P2P)用户习惯、运行环境的异构性，提出P2P流量识别的双层模型。该模型由单流内部流量特征的贝叶斯网络识别算法与多流之间行为特征的支持向量机识别算法组成。实验结果表明，相对于统计特征识别方法，该模型检测准确度提高5.4%，且对于不同应用场景具有较好的稳定性。

关键词: 流量识别, 点对点, 双层模型, 贝叶斯网络, 支持向量机, 行为特征

Abstract: Considering the heterogeneity of Peer-to-Peer(P2P) users habit and runtime environment, this paper proposes a two-layered model of P2P traffic identification to identify and filter the P2P traffic. It combines both an identification algorithm of Bayesian network based on single traffic feature method and an identification algorithm of Support Vector Machine(SVM) based on multi-traffic behavior method. Experimental results show that the method is 5.4% more accurate than the statistical feature identify method, and it has better stability in different application scenes.

Key words: traffic identification, Peer-to-Peer(P2P), two-layered model, Bayesian network, Support Vector Machine(SVM), behavior feature

中图分类号: 

• TP393.08