Reproduce “IVE: An Accelerator for Single-Server Private Information Retrieval Using Versatile Processing Elements” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sangpyo Kim, Hyesung Ji, Jongmin Kim, Wonseok Choi, Jaiyoung Park, Jung Ho Ahn
An accelerator for single-server homomorphic encryption (HE)-based Private Information Retrieval (PIR), named IVE, provides up to 1,275x higher throughput and 51.3x lower energy consumption per query compared to prior hardware, enabling practical private queries for terabyte-scale databases. It integrates versatile processing elements and hierarchical data reuse algorithms to address computational and memory bottlenecks.
Key insights
- A versatile systolic NTT unit (sysNTTU) that combines NTT and GEMM functions within a single hardware unit improves area efficiency for sequentially dependent cryptographic workloads.
- Hierarchical Search and Reduction Overlapping effectively reduce DRAM data transfers by maximizing on-chip data reuse and minimizing temporary memory footprint for client-specific HE computations.
- Integrating heterogeneous memory and a scalable cluster architecture enables practical deployment of single-server HE-PIR for databases ranging from gigabytes to terabytes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.