Reproduce “The Future of Fully Homomorphic Encryption System: from a Storage I/O Perspective” 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
Lei Chen, Erci Xu, Yiming Sun, Shengyu Fan, Xianglong Deng, Guiming Shi, Guang Fan, Liang Kong and 3 more
This research quantifies the impact of storage I/O on Fully Homomorphic Encryption (FHE) system performance, revealing that data movement, rather than computation alone, frequently bottlenecks FHE accelerators in cloud environments, leading to significant performance degradation.
Key insights
- Storage I/O access profoundly degrades FHE application performance, with slowdowns reaching hundreds of times compared to ideal memory-resident baselines, establishing I/O as a primary bottleneck.
- Employing multiple servers for distributed computing does not consistently resolve performance issues and can introduce communication overheads, sometimes even harming performance for certain accelerator types like Sharp ASICs.
- The impact of I/O overhead is highly variable, differing between applications (e.g., HELR vs. ResNet-20) and within the same application depending on the FHE parameters chosen, which influence ciphertext and key sizes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.