AI & ML Breaks Assumption

Develops a zero-watermarking framework that survives AI editing by leveraging invariant relations between image patches.

March 19, 2026

Original Paper

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

Pengzhen Chen, Yanwei Liu, Xiaoyan Gu, Xiaojun Chen, Wu Liu, Weiping Wang

arXiv · 2603.17531

The Takeaway

Challenges the assumption that watermarks must be embedded into pixel data, which is easily erased by diffusion-based editing. By using the 'relational distance' between patches, it provides a non-invasive way to authenticate content that remains robust even after heavy AI-driven modifications.

From the abstract

Recent advancements in diffusion-based image editing pose a significant threat to the authenticity of digital visual content. Traditional embedding-based watermarking methods often introduce perceptible perturbations to maintain robustness, inevitably compromising visual fidelity. Meanwhile, existing zero-watermarking approaches, typically relying on global image features, struggle to withstand sophisticated manipulations. In this work, we uncover a key observation: while individual image patche