attack arXiv Mar 1, 2026 · 5w ago
Huajie Chen, Tianqing Zhu, Hailin Yang et al. · City University of Macau · CISPA Helmholtz Center for Information Security +1 more
Pixel-wise reconstruction attack removes AI-image watermarks without querying detectors or knowing the watermarking scheme
Output Integrity Attack visiongenerative
Watermarking has emerged as a key defense against the misuse of machine-generated images (MGIs). Yet the robustness of these protections remains underexplored. To reveal the limits of SOTA proactive image watermarking defenses, we propose HIDE&SEEK (HS), a suite of versatile and cost-effective attacks that reliably remove embedded watermarks while preserving high visual fidelity.
diffusion gan City University of Macau · CISPA Helmholtz Center for Information Security · Guangzhou University
ML Security Papers