Statistical deanonymisation

Hello lainons, I watched the DEF CON 25 talk "dark data" and I wonder how the risks to tor users could be mitigated. 1- Would accessing links that you have only accessed a few times without tor be a risk? 2- Would months or years between the access times without tor and with tor reduce the risk? 3- Would using qubes and individual whonix instances for each link reduce the risk of fingerprinting? Thank you.

>>21891 The biggest risk is if you use the same browser with a unique fingerprint to visit a site with and without Tor. That's why the torbutton extension was killed and replaced with tor browser.

>>21891 1 - obviously 2 - i think it depends on how active that site is and what kind of website it is. If it's a forum or imageboard and you post on it, you might have a higher risk of being identified by stylometry. 3 - should almost entirely eliminate the risk of fingerprinting as long as you don't compromise yourself, since you now have a completely homogeneous system as long as you keep it stock. Don't change the screen resolution, don't install unnecessary fonts, don't change the system timezone etc.

This is the OP, thank you for the responses so far. Just to clarify, Is there a risk of being deanonymised due to visiting(but not posting on) a possibly unique set of websites that I previously accessed without tor, or is this only an issue if it links back to a social media profile or website with my personal information included. I saw this mentioned on the anonymousplanet.org guide. Thank you.

>>21895 As long as you're using a different tor circuit for each connection (I think this is default on the for browser now) then you're fine in this regard.

>>21895 Yes, there is a risk.

>>21901 Do you think this would still be a risk when visiting the website's onion address, if available, or do you think they would still be linked? Thank you.

Bump, write a longer post error.

>>21930 Lainchan is relatively slow and people check the catalog, you really don't need to uselessly bump a thread, especially after just 4 days since its last post. >>21895 >>21902 It depends on a lot of things. Things that will weaken Tor's anonymity would include accessing the site via URLs that may be uniquely accessed by you, accessing the site from the same browser as you accessed it without Tor in the past, and visiting a unique set of websites at the same time (especially low traffic sites). The latter two are mostly mitigated just by using the Tor browser bundle, which tries to use one circuit per "site" (not literally one circuit per site, since if you go to site A and it embeds something from site B, it's ideal if both requests are on the same circuit). There are still attacks possible, but really the risk is minimal unless your threat model needs to account for being targeted by serious adversaries. You're not going to be deanonymized because you visited dark.fail without tor a few months ago or anything like that. There are good Tor related papers here: https://www.freehaven.net/anonbib/ https://research.torproject.org/techreports/