摘要： 针对当前基于流特征的流量识别方法准确率较低的问题，提出一种基于信息熵的流量识别方法，运用信息熵寻找显著特征，根据显著特征进行级联分簇。实验分析表明，该方法识别流和字节的准确率达90%以上，比单纯用K-Means等聚类算法的准确率提高10% 左右。

关键词: 流量识别, 信息熵, K-Means算法, 深度包探测

Abstract: In order to improve accuracy of identification traffic based on flow features, a new method of Traffic Identification Based on Information Entropy(TIBIE), which uses information entropy to find the notable features, and cascades clustering according to the notables features are put forward. Experimental results show that both the flow and byte accuracy rate of this method achieve 90% or more, which increase about 10% than the method that only based on K-Means clustering algorithm.

Key words: traffic identification, information entropy, K-Means algorithm, Deep Packet Inspecting(DPI)

中图分类号: 

• TP393.08