Reproduce “Location Matters: Sparse Latent-Variance Attacks for Face Privacy against Diffusion Personalization” 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
Shuting Yang, Shuxin Li, Linyan Dai, Xinwei Zhang, Dong Huang, Renyang Liu, See-Kiong Ng
SEAL protects public face photos from fine-tuning-based diffusion personalization by optimizing a sparse set of perturbed pixel locations, using latent-variance disruption and a facial-saliency penalty. On VGGFace2 with Stable Diffusion 2.1 and DreamBooth, it modifies 5% of locations and reports PSNR 41.08 dB, SSIM 0.97, LPIPS 0.07, and ISM 0.15 under P2. It does not protect encoder-based personalization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.