基于优化SVM的P2P协议识别  

Novel P2P protocol identification algorithm based on optimized SVM

作　　者：毛灵[1] 陈兴蜀[1] 吴仲光[1] 谭骏[1] 杜敏[1] 

机构地区：[1]四川大学计算机学院,成都610065

出　　处：《计算机应用研究》2011年第7期2750-2753,共4页Application Research of Computers

基　　金：国家"973"计划重点基础研究发展资助项目(2007CB311106);国防重点实验室基金资助项目(NEUL20090101)

摘　　要：针对P2P应用提出了一种采用DFI深度流分析的方法,通过还原会话流,提取P2P数据流的各种属性特征,采用Grid Search、遗传算法、粒子群算法三种不同算法优化的支持向量机对网络数据流进行分类。通过实验测试,在P2P与非P2P的多种应用中,使用支持向量机进行设计的分类器分类准确率较高,均在90%以上,最高能达到97%。To overcome the shortcomings of P2P application,this paper proposed a new P2P identification algorithm which used DFI deep flow analysis by extracting the various properties of P2P data stream characteristics,using support vector machine which was optimized by Grid Search,generation algorithm and particle swarm optimization to assort the network data streams.Experimental results of the P2P and non-P2P classification show that the classifier by SVM has a high accuracy over 90%,and the highest can be 97%.

关 键 词：粒子群算法 遗传算法 支持向量机 P2P 协议识别 

分 类 号：TP393.0[自动化与计算机技术—计算机应用技术]