一种基于混沌粒子群算法和支持向量机的P2P流量识别方法

A PEER TO PEER TRAFFIC IDENTIFICATION METHOD BASED ON CHAOS PSO AND SVM

对等网络技术P2P(Peer-to-Peer)在丰富了互联网应用的同时也带来了很多安全问题,因此,如何进行P2P流量的识别是网络管理研究的热点和难点问题。其中支持向量机在P2P识别问题中具有较好的效果,然而支持向量机的分类性能很大程度取决于核函数参数和惩罚参数。基于遗传算法、粒子群算法的支持向量机参数优化方法都存在易陷入局部最优解的问题,优化性能需要进一步改善。为进一步改善支持向量机参数优化问题,提出一种基于混沌粒子群的支持向量机参数优化方法,并将其应用于P2P流量识别问题。利用真实的校园网网络流量作为研究对象进行分类实验,结果表明,混沌粒子群优化的支持向量机具有更高的P2P分类正确率和计算效率。

Peer-to-peer（ P2P） network technology brings many security problems while enriching the applications of Internet. Therefore the P2 P traffic identification is the hot topic and difficulty in networks management research. Support vector machine（ SVM） has preferable effect in P2 P identification,but its classification performance is largely dependent on kernel function parameter and punishment parameter.And the SVM parameters optimisation approaches based on genetic algorithm and particle swarm optimisation（ PSO） all have the problem of easily falling into local optima,their optimisation performances require further improvement. In order to further improve this performances optimisation issue,we proposed a chaos PSO-based SVM parameters optimisation method and applied it to P2 P traffic identification. We used traffic data in actual campus networks as the research object to carry out classification experiment. Results showed that the support vector machine optimised by chaos particle swarm optimisation had higher P2 P classification accuracy and computation efficiency.