Key Technologies And Research Of Anonymouns Website Fingerprinting Attack

With the development of the Internet,people can visit websites and browse web pages anytime and anywhere with the use of Web technology,which greatly facilitates the acquisition and interaction of information and accelerates the development of human society in the information age.While facilitating people’s lives,browsing also exposes users’ privacy.In order to protect user privacy,anonymous network technologies have made great progress.From early SSH proxy technology to more sophisticated and advanced Tor anonymous network,people’s personal communication privacies have been better protected.However,any technology is a double-edged sword,and anonymous network technology is no exception.Criminals use anonymous networks to hide their illegal activities,thus escaping legal sanctions,which brings great difficulties to network monitoring and management,and seriously endangers the security of network space in China.In order to better regulate anonymous networks,investigate and collect evidence of network crimes,and maintain the security of national network space,this paper takes the key technologies of website fingerprinting attack in anonymous networks as the research topic,focusing on the existing methods and limitations.On this basis,the following key problems are studied: How to use only a small number of samples from new websites for few-shot website fingerprinting attack without additional data? How to effectively extract knowledge from additional data and migrate it to the few-shot website fingerprinting attack when there is a large amount of additional data? How to design the training task as same as the target test task,so that the few-shot website fingerprinting attack model can better adapt to the current specific task? Facing the fact that users tend to browse multiple pages with multiple tabs at the same time,how to develop a more realistic multi-tab website fingerprinting attack method.The major work and innovations of this thesis are as follows:In the first part,a new few-shot website fingerprinting attack with data augmentation is presented.This study proposed a novel data augmentation menthod for few-shot website fingerprinting attack where only a handful of training samples per website are available for deep learning model optimization.In this study,besides three proposed augmentation operations,a harmonious data augmentation algorithm is proposed so that the proposed operations can be used in a harmonious manner.The experimental results show that the performance of the start-of-tha-art deep learning based website fingerprinting attack algorithm can be significantly improved in different scenarios.In the second part,based on knowledge transfer,a novel Transfer Learning Fingerprinting Attack(TLFA)is presented.This study aimed at the situation that some monitoring tasks have a large amount of additional data available,but the existing few-shot website fingerprinting attack technology can not effectively used them.TLFA can transfer knowledge from the labeled training data of websites disjoint and independent to the target websites.We conduct expensive experiments to validate the superiority of our TLFA over the state-of-the-art methods in both closed-world and open-world attacking scenarios,at the absence and presence of strong defense.In the third part,a novel Few-shot Website Fingerprinting Attack with Meta-Bias Learning(MBL)is proposed.Taking the meta-learning strategy,MBL simulates and optimizes the target tasks.Moreover,a new model parameter factorization idea is introduced for facilitating meta-training with superior task adaptation.Expensive experiments show that our MBL outperforms significantly existing hand-crafted feature and deep learning based alternatives in both closed-world and open-world attack scenarios,at the absence and presence of defense.In the fourth part,a novel end-to-end multi-tab website fingerprinting attack is proposed.Indeed,this multi-tab website fingerprinting attack setting has been studied in a few recent works,but all of them still fail to fully respect the real-world situations.we formulate a novel Website Fingerprint Detection(WFD)model capable of detecting accurately the start and end points of all the monitored traces and classifying them jointly,given long,untrimmed raw traffic data.Extensive experiments on several newly constructed benchmarks show that our WFD outperforms the state-of-the-art alternative methods in both accuracy and efficiency by a large margin,even with a very small training set.