基于BPSO与神经网络的实时P2P协议识别算法

针对互联网中P2P协议以及加密协议无法使用传统方法进行识别的问题,提出一种新的基于会话流统计特征的网络协议识别算法。采用二进制粒子群算法(BPSO)定量选出最能体现不同协议区别的特征子集;并针对BP(Back Propagation)神经网络结构难以确定、易陷入局部极小值等缺陷进行分析,使用粒子群算法对BP神经网络进行优化以提高识别率。实验结果表明:该方法能够有效地从多种网络特征属性中选出最能体现不同协议区别的特征子集,且对于基于UDP协议的网络应用也有较高识别率,经优化后的BP神经网络具有更高识别率。该算法对常见的P2P协议平均识别率达到96%,且能够实时地对网络协议进行识别。

A novel real-time p2p identification algorithm based on BPSO and neural networks

Due to the unclassifiable problem of P2P protocol and encryption protocol by traditional approach in network management,a novel approach considering internet traffic flow was proposed to classify network applications especially P2P applications based on binary particle swarm optimization(BPSO) and optimized back-propagation(BP) neural network.BPSO was used to select the best feature subset which can mostly reflect the difference among different network applications.And BP neural network was optimized by PSO algorithm.The experimental results demonstrate that the proposed approach has a high recognition rate of network applications using either TCP or UDP protocol,and the identification rate is improved to 96% with the use of BPSO and optimized BP neural network.Moreover,the proposed algorithm can be used for real-time identification.