Rosemary paper accelerates FHE with in-storage SSD processing
Original title: Rosemary: Efficient and Adaptable In-Storage Acceleration for Fully Homomorphic Encryption
The design reports 185.7× geometric-mean speedups over CPU and GPU FHE systems on encrypted workloads.
A cs.CR paper presents Rosemary, a heterogeneous fully homomorphic encryption system that moves FHE work into a NAND flash SSD to cut storage I/O. The authors say current FHE systems spend 71% to 85% of runtime moving data rather than computing. Rosemary combines an in-drive FHE engine, a workload allocator and a ciphertext-aware compiler optimizer, and is evaluated on encrypted deep-learning inference and cloud database queries.
- Reported speedups are 185.7× over CPU/GPU systems and 3.7× over prior FHE accelerators.
- Energy efficiency improves by up to 3.9× and EDAP by up to 16.5×.
- The hardware target is a NAND flash-based solid-state drive.
- Products
- Rosemary
From the original article
1 min read at arxiv.orgFully Homomorphic Encryption (FHE) enables computation on encrypted data and is promising for privacy-preserving computing. However, FHE deployment remains constrained due to massive computational, memory, and data-movement overheads.
By Xuan Wang · 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.