HPS-PEC uses register renaming for privacy-enhanced CPU computing
Original title: Hidden in Plain Sight: Privacy-Enhanced Computing Through Register Renaming
Across 25 VIP-Bench benchmarks, encrypted code ran 3.41x native time and 1.26x faster than a sequestered unit.
Researchers propose HPS-PEC, which reuses existing CPU register-renaming infrastructure for privacy-enhanced computing instead of a dedicated processor enclave. The design aims to keep plaintext unreachable by every instruction, assuming a correctly functioning renamer and register file. The authors say this avoids duplicate ALUs and allows out-of-order and parallel PEC execution.
Unconfirmed: Speed and area figures are the authors' measurements, not independently verified.
- Evaluation uses a 7-cycle QARMA-64 cipher.
- Data-oblivious code itself accounts for a 1.03x slowdown.
- HPS-PEC uses under a third of the sequestered unit's dedicated area.
- Prior PEC designs require a dedicated enclave and duplicate ALU resources.
From the original article
1 min read at arxiv.orgPrivacy-Enhanced Computing (PEC) is a mechanism to hide program secrets from the outside world without paying the cost of fully-homomorphic encryption.
By Ahmed Nematallah · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.