[tor-talk] Practical deanonymization using CPU load covert channels

[bancfc at openmailbox.org]

Thanks Ethan for your research. We've been discussing which mitigation 
technique to use. In a virtualized context disabling c-states is only 
possible from the host. Thats ok since all it means is we need to 
package it for users to install it there.

We prefer the idea of not using the kernel commandline directly because 
of discrepancies between hypervisors and custom kernels. We are leaning 
towards using the pmqos-static.py script some_guy123 mentioned instead.


Please feel free to jump into the discussion on our bugtracker any time:

https://phabricator.whonix.org/T530