Piercing the cloud and marking the onion: Removing the anonymity of the Tor network

In this work, we extend previous work done on removing anonymity from The Onion Routing network (Tor). We explore previous techniques for removing Tor's anonymity developed on a private Tor network, and attempt to reproduce these results on the global public network.;We find that the previous work done on the private network is unable to be carried over to the public network. This is mainly due to the level of jitter on the public network overwhelming our earlier method for compromising Tor's anonymity. We develop a new method for compromising Tor's anonymity by using a clustering algorithm to analyze the data that we gathered via a side channel timing attack. This neural network finds data clusters that can be recognized despite the jitter in the global network. We then used use the recognizable timing patterns to build Hidden Markov models (HMM). Using these models we are able to recognize network traffic patterns and reduce Tor's anonymity.;We establish how well multiple paths through Tor prevents our side channel attack. Because the paths don't contain the same nodes, the packet delays are different. This successfully counters our side channel attack and restores Tor anonymity.