You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The main idea to mitigate the kind of attacks described in our threat model (Section no?) is to hide the information about the set index from an adversary. This ties in well to the philosophy of a fully associative cache where we have a single set allowing us to place the block in any of the ways. Therefore, any knowledge obtained from the eviction of attacker's cachelines does not provide him with any meaningful information about the victim's address bits. We try to achieve this by introducing randomness in the placement of blocks in the cache. This effectively increases the associativity of the cache as was shown by Seznec et al in \cite{skewed cache}. It gives us an additional benefit of reduced conflict misses.