Reproduce “Enabling Practical and Privacy-Preserving Image Processing” 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

Chao Wang, Shubing Yang, Xiaoyan Sun, Jun Dai, Dongfang Zhao

Researchers at Worcester Polytechnic Institute and the University of Washington developed iCHEETAH, a pixel-level homomorphic encryption system built on the CKKS scheme, which significantly improves the practicality of privacy-preserving image processing. It achieves up to a 19-fold speedup in encryption through caching mechanisms while maintaining IND-CPA security and high image quality, enabling various operations on encrypted images up to 1024x1024 pixels.

Key insights

  • Optimized caching strategies can significantly mitigate the computational overhead of FHE for image data, making pixel-level homomorphic encryption practical for images up to 1024x1024 resolution.
  • The integration of randomness to ensure IND-CPA security can be achieved with minimal performance impact, demonstrating that efficiency gains from caching are not negated by security requirements.
  • Applying fundamental image processing operations like filtering, brightness adjustment, and watermarking directly on pixel-level encrypted data is feasible and retains expected outcomes, opening avenues for secure cloud-based image services.

Summary by alphaXiv.

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