The Construction and Experimental Approach of Anonymous Network Analysis and Control Platform

According to the traffic features of anonymous networks such as TOR and I2P, this paper proposes an anonymous network traffic filtering approach based on payload size entropy. It exploits the filtered traffic as the training set to build a machine learning model for traffic classification, by adopting decision tree learning algorithm. In order to extract target attributes from multiple characteristic attributes of anonymous network traffics, the Bayesian classification algorithm is utilized to analyze the impact of each attribute on the information entropy gain, so as to select target attributes in the traffic classification model. To be specific, this paper focuses on constructing an anonymous network analysis and control platform, and carries out experimental verification in an anonymous network environment. The experimental results show that the identification accuracy of the proposed approach for anonymous network traffic is as high as 96.8%, and the packet loss rate is less than 0.1% in the Gigabit network environment within the allowable error range.