Adaptive Obfuscation System Based On Meek For Antitraffic Analysis

With the rapid development of Internet technology,the technology used for tracking and monitoring is also constantly improving.When facing various network security issues,people tend to use anonymous communication technology to protect privacy.Tor is one of the most widely used anonymous communication technologies known.To facilitate users to successfully break through censorship,Meek is introduced as a pluggable transport,and the Meek-pluggable transport exists in the process of disguising Tor traffic.The problem of privacy exposure,on the one hand,when faced with machine learning-based traffic analysis attacks,the statistical characteristics of traffic make Meek easy to be distinguished.On the other hand,the traffic features exposed by the Meek forwarding mechanism reduces the difficulty of attackers' traffic analysis.At the same time,anonymous user groups can be easily traced.Aiming at the privacy leakage of Tor anonymous communication system users based on the Meek-pluggable transport,the thesis proposes an adaptive traffic obfuscation method,which de-differentiates the side channel features of traffic packets,and introduces SDN architecture,proposes an SDN adaptive obfuscation scheme based on node filtering,which introduces Meek traffic into obfuscated paths that are difficult to trace,and at the same time,it can hide packet information.The specific research content is divided into three aspects:(1)An adaptive traffic obfuscation scheme based on side-channel features is proposed.The scheme collects traffic for potential passive attackers to analyze the communication relationship between the communication parties and uses nonlinear programming to change the traffic features.This scheme aims to minimize the deformation cost,filter out suitable target traffic through clustering and cosine similarity methods,and disguise the features of sensitive traffic as ordinary traffic,thereby confusing attackers and increasing identification costs.Finally,the clustering effect of the scheme was demonstrated,and its effectiveness was verified by three machine learning models.(2)A SDN adaptive obfuscation scheme based on node screening is proposed.This scheme uses the idea of self-adaptation and randomization,and the process can be divided into two parts: node self-adaptive screening and multi-path calculation.The various processes cooperate to screen out the appropriate node-set and path,and when transmitting sensitive information,randomly select the transmission path from the set for the flow data packet.When targeting malicious nodes,use global credibility to filter nodes,and update the set of trusted nodes based on this.Finally,the anonymity and security of this method are analyzed.(3)We use Mininet tools,Ryu controller,etc.to build an SDN network simulation platform,design and implement an adaptive obfuscation prototype system.The system adopts a structured programming concept to realize an adaptive traffic obfuscation scheme based on side-channel features.Based on the Mininet simulation environment,the connectivity of each node under the system,node filtering,and path calculation functions was tested,and the path information was displayed through the Web interface.Finally,the traffic obfuscation effect was tested through the WEKA tool.