| The popularity of the Internet has facilitated people’s lives,but it has also brought convenience to crime.The network that people usually come into contact with is the Internet and is generally presented by the World Wide Web.The World Wide Web is a web page interconnection and expression system based on hyperlinks and hypertext languages.There is a kind of data outside of this,which is built on the Internet,but it is not concerned.This type of data is stored in structured data and most requires authentication to access it.It cannot be acquired by search engines and cannot be retrieved by the general public,that is,"deep network." In the "deep network" data,some of the data generated by completely anonymous users,that is," Tor network"."Tor network" is a multi-layer encryption and multi-node access technology under the concept of "peer-to-peer network",organized into an information exchange network,and its biggest feature is complete anonymity."Tor network" is widely used because of its anonymity,and it has become a paradise for criminals,which has brought certain difficulties to criminal evidence collection.Therefore,how to break the anonymity in the "Tor network" to achieve tracking is very important.In the existing methods,the we bsite fingerprinting attack method is one of the more popular research fields,and new knowledge is continuously introduced to make the method obtain better accuracy.However,there are certain deficiencies,such as single consideration.The web page does not take into account the interaction of the site.In this paper,an anonymous network in the "tor network" is selected as the research object,and the fingerprint ing attack method is used to realize information tracking.The main research work includes:1.Introduced the shortcomings of anonymous network,reviewed the current status of website fingerprint attack methods at home and abroad,summarized the anonymous network-based attack methods and made a comparative analysis.2.Studied site fingerprint attack wavelet transform of the collected data wavelet transform process,using the SVM algorithm classifies generated fingerprint database,by experimental screening suitable wavelet basis and packet stuffing length,a 10-fold cross-validation method verified the accuracy of the method.Simulation experiments were carried out on the data sets DS1 and DS2,and the experimental results were analyzed.3.Studied the website fingerprint attack method based on Hidden Markov Model,generated the fingerprint feature into a feature sequence,and used the gene sequence comparison method to detect its accuracy.The feasibility analyzed of this method was carried out.4.Studied the fingerprint attack method of SSH anonymous network website,applied Adaptive Hoeffding Tree and Adaptive Hoeffding Option Tree algorithms,and adopted Test-Then-Train method to verify the accuracy of the method.Simulation experiments were carried out on the data sets D1,D2,D3,and the experimental results were analyzed.The innovation of this paper includes:1.A method of website fingerprinting attack based on wavelet transform was proposed.The fingerprint information is used to identify and track website information.Experiments on data sets DS1 and DS2 showed that this method had better accuracy than the website fingerprinting attack method based on Fast Fourier Transform2.A method of website fingerprinting attack based on Hidden Markov Model was proposed.This method realized the forensic website access on the Tor anonymous network,taking into account the impact of accessing other address links through the website when visiting the website.A serialization algorithm was proposed to make this method suitable for TOR anonymous networks.3.A website fingerprint attack method based on SSH anonymous network was proposed,which realizes website access forensics on SSH anonymous network,considering the impact of time behavior of dynamic websites on forensics.Experiments on data sets D1,D2,D3 showed that this method could reduce the impact of temporal behavior on forensics,and had a better accuracy than the fingerprint attack method of websites based on convolution neural networks. |
