Seeds of SEED: HCache: Building a Hybrid Randomized Cache Hierarchy for Mitigating Cache Side-Channel Attacks

Xingjian Zhang, Ziqi Yuan, Rui Chang, Yajin Zhou

Abstract: Cache side-channel attacks can leak critical information from the target programs. The cache randomization methodology has proven to be an efficient way to mitigate such attacks. However, existing works do not take the cache hierarchy into consideration, failing to address the issue that different levels of caches have different performance and security requirements. In this work, we propose and implement a hybrid randomization scheme, named H2Cache, to mitigate cache side-channel attacks. H2Cache leverages two randomization approaches and applies them to different levels of caches. It strengthens the security of cache modules, while satisfying the performance and resource utilization requirements. Specifically, we design a table-based randomization method for the L1 cache, which uses a hashed virtual index to look up the actual cache set index. The L2 cache in H2Cache takes a computation-based randomization function to calculate the cache set index. We have implemented a prototype of H2Cache and extensively evaluated it using a self-designed RISC-V processor on the FPGA platform. We demonstrate the security of H2Cache through simulated attack programs and quantitative analysis. Meanwhile, the evaluation results of performance and resource utilization have shown its efficacy.