A Transfer Attack to Image Watermarks

arXiv:2403.15365v1 Announce Type: new
Abstract: Watermark has been widely deployed by industry to detect AI-generated images. The robustness of such watermark-based detector against evasion attacks in the white-box and black-box settings is well understood in the literature. However, the robustness in the no-box setting is much less understood. In particular, multiple studies claimed that image watermark is robust in such setting. In this work, we propose a new transfer evasion attack to image watermark in the no-box setting. Our transfer …

arxiv attack attacks box cs.cl cs.cr cs.lg detect detector evasion evasion attacks generated image images industry literature robustness settings studies transfer watermarks