Re:Passive Deanonymization of Tor Hidden Services [pdf]
by peterwwillis on Thursday December 03, 2015 @10:42AM
(#10670156)
In general, it depends. Statistical correlation analysis works pretty well as long as you have enough data, the data points you're looking at aren't specifically designed to change constantly (such as with a strong boolean function in a stream cipher correlation attack), and they don't introduce false positives. But if the data points you're looking at are strong indicators of unique data and your training method can correlate strongly with a specific user's requests, it shouldn't be too difficult.
Section 7 (page 12) goes over website fingerprinting and how they're able to gain a significant positive match and low false-positive match rate. They can more easily identify noise from signal than previously thought, we're only looking at hidden services (which number far fewer than the number of public websites), hidden service pages change much less frequently than public ones, and they exclude data points that could raise the false positive rate.
Also note that the classifiers here were even better at identifying specific clients than what hidden services they were accessing. So, not only is your hidden service not safe, neither are your users :-)
Section 7 (page 12) goes over website fingerprinting and how they're able to gain a significant positive match and low false-positive match rate. They can more easily identify noise from signal than previously thought, we're only looking at hidden services (which number far fewer than the number of public websites), hidden service pages change much less frequently than public ones, and they exclude data points that could raise the false positive rate.
Also note that the classifiers here were even better at identifying specific clients than what hidden services they were accessing. So, not only is your hidden service not safe, neither are your users :-)