基于PCA的对等网络流量时空特性监测

PCA based approach for monitoring the spatial-and-temporal characteristics of P2P traffic

在对等网络中,节点可以动态地进入和离开网络,增加了业务监测的难度。该文提出了采用主成分分析(PCA)方法检测并分析对等网络流量特征,解释了流量协方差矩阵最大特征值和最大特征矢量的物理意义,同时定义权重矢量作为流量观测指标,利用流量的协方差矩阵最大特征值和特征向量给出了全网的P2P流量时间和空间的动态特性。仿真结果表明,该方法能够很好地识别P2P流量的动态时间和空间特性,为互联网流量检测提供了一种有效方法。

Monitoring P2P traffic is difficult due to the peers in P2P networks joining and leaving dynamically. This paper presents a method based on principal components analysis （PCA） to monitor the spatial-and-temporal characteristics of P2P traffic. The meaning of the largest eigenvalue and the corresponding eigenvector is explained by the PCA theory,which are calculated from the cross-correlation matrix of the flow data. A weight vector is also defined to monitor traffic patterns. Simulation results show that the spatial-and-temporal characteristics are effectively captured using the developed method.