It has long been known that timing analyses are a *theoretical* attack on Tor. By distributing the circuits across different jurisdictions, the goal was to make these attacks impractical to implement:
Only a "global adversary" should be able to break the anonymity by correlating the traffic from entry and exit nodes. Correlation becomes even easier if delays or content can be actively introduced into the traffic pattern.
Just as we could (theoretically) become a "global adversary" by renting enough servers, law enforcement agencies can (practically) achieve this through close cooperation, especially since Tor nodes are not evenly distributed across jurisdictions but tend to cluster in certain regions.
Western law enforcement agencies seem to have reached the "global adversary" level through collaboration (in isolated cases and certainly with significant effort). What is problematic for Tor is that other "law enforcement agencies," whose focus is on dissidents, whistleblowers, and journalists, could do the same.
So, it is finally time for cover traffic and random delays: nodes in the Tor network would introduce a random traffic background noise as well as random delays to make targeted correlations more difficult. This would make Tor even slower. This is probably why it has been avoided until now.
In conclusion, we would like to emphasize that there is no reason for regular users of the Tor browser to worry about their anonymity. These are highly targeted attacks on individual accounts of the messenger "Ricochet" over extended periods of time. Because the messenger, unlike a browser, is also reachable, it naturally has an increased attack surface for timing analyses.
https://www.tagesschau.de/investigativ/panorama/tor-netzwerk-100.html
