基于混合特征的P2P流量识别方法

Hybrid Characteristics Based P2P Traffic Identification Methods

研究P2P异常流量的识别问题。P2P网络节点特征属性较多,代表流量特征的属性存在多层属性,传统的流量识别方法以整体特征为基础,没有对流量特征属性进行进一步划分,一旦出现多识别特征的情况,单一类内的特征很难准确描述这种多流量特征,导致识别精度下降。为了避免上述传统算法的缺陷,提出了一种基于支持向量机增量学习算法的p2p流量识别方法。提取p2p流量混合特征,并将其作为p2p流量识别的依据。建立支持向量机增量学习模型,并对提取的流量混合特征进行有效的识别。实验结果表明,利用改进后的算法能够对异常流量进行准确的识别,提高异常流量识别率,降低误判率,从而有利于p2p网络的管理。

The identification problem of P2P abnormal traffic was studied in this paper. We proposed a p2p traffic identification method based on SVM incremental learning algorithm. Firstly, p2p traffic mix features were extracted to used as the basis of p2p traffic identification. Then a support vector machine incremental learning model was created, and the extracted traffic hybrid feature can be effectively identified. Experimental results show that taking the advan- tage of the improved algorithm can identify the abnormal traffic accurately, increase the recognition rate of abnormal traffic and reduce false positives, thus contributing to p2p network management.