Reproduce “OptiPrime: Optimizing Private Inference through Protocol-Hardware Co-design” 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

Jiangrui Yu, Ye Yu, Si Chen, Chenqi Lin, Wenxuan Zeng, Junfeng Fan, Mingyu Gao, Meng Li

OptiPrime improves two-party private neural-network inference by packing convolution results into fewer ciphertexts and redesigning server memory handling, while homomorphic encryption processes linear layers and multiparty computation handles nonlinearities. In a specified ResNet-18 layer, it uses 8 output ciphertexts instead of Cheetah’s 256. Across evaluated ImageNet settings, latency falls by up to 5.7× on CPUs and 4.2× with an accelerator.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.