Rel-Zero: Harnessing Patch-Pair Invariance for Robust Zero-Watermarking Against AI Editing

Rel-Zero introduces a novel zero-watermarking framework that utilizes the invariance of relational distances between image patch pairs to authenticate digital content without altering the original image. This method circumvents the perceptual issues associated with traditional watermarking by grounding the watermark in structural consistency, achieving enhanced robustness against various AI-based editing techniques. Its significance lies in providing a non-invasive solution for content authenticity in an era increasingly challenged by advanced image manipulation methods.