Reproduce “Chameleon: An Efficient FHE Scheme Switching Acceleration on GPUs” 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
Zhiwei Wang, Haoqi He, Lutan Zhao, Peinan Li, Zhihao Li, Dan Meng, Rui Hou
Chameleon is a GPU-accelerated framework designed to enhance the efficiency of Fully Homomorphic Encryption (FHE) scheme switching, critical for hybrid FHE applications. It optimizes Number Theoretic Transform (NTT) operations for diverse polynomial sizes and accelerates key conversion processes like Look-Up Table (LUT) evaluation and repack, achieving substantial speedups up to 1.58x for NTT and an average of 67.3x for overall scheme switching over prior implementations.
Key insights
- Intelligently fusing NTT stages and rearranging data layout to ensure interdependent coefficients reside within the same thread block can drastically reduce GPU synchronization overhead across varying polynomial sizes.
- Exploiting data-level parallelism at the CMux gate level, rather than individual ciphertext processing, significantly improves the efficiency of TFHE's LUT evaluation by reducing redundant operations and increasing GPU utilization.
- Algorithmic modifications, such as merging automorph operations with linear transformations, can eliminate the need for expensive homomorphic rotations during FHE scheme conversion, leading to substantial performance gains.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.