Cache-Timing Attacks and Shared Contexts

Abstract (English)

Cache-timing attacks recover algorithm state by exploiting the fact that the latency of retrieving data from memory is essentially governed by the availability of said data in the processor’s cache. Efficient and effective countermeasures to these attacks are needed. A shared memory context is a mechanism for reusing dynamically allocated memory. Focusing on public key cryptography within OpenSSL and its implementation of shared contexts, this paper examines the ability of a shared context to aid in mitigation of cache-timing attacks. The results are pessimistic towards this approach.