Search papers, labs, and topics across Lattice.
The paper introduces ShapeMark, a novel watermarking technique for diffusion models that improves robustness and preserves generation diversity. ShapeMark encodes watermark bits into structured noise patterns rather than individual values, making it more resilient to perturbations. To maintain diversity, the method randomizes the positions of noise elements without altering their values, preventing fixed noise patterns. Experiments show ShapeMark achieves state-of-the-art robustness and high generation quality under various lossy conditions.
ShapeMark watermarks survive heavy image degradation by encoding bits into structured noise patterns, unlike existing methods that embed in individual pixel values.
Diffusion models have made substantial advances in recent years, enabling high-quality image synthesis; however, the widespread dissemination and reuse of their outputs have introduced new challenges in intellectual property protection and content provenance. Image watermarking offers a solution to these challenges, and recent work has increasingly explored Noise-as-Watermark (NaW) approaches that integrate watermarking directly into the diffusion process. However, existing NaW methods fail to balance robustness and diversity. We attribute this weakness to value encoding, which encodes watermark bits into individual sampled values. It is extremely fragile in practical application scenarios. To address this, we encode watermark bits into the structured noise pattern, so that the watermark is preserved even when individual values are perturbed. To further ensure generation diversity, we introduce a dedicated randomization design that reshuffles the positions of noise elements without changing their values, preventing the watermark from inducing fixed noise patterns or spatial locations. Extensive experiments demonstrate that our method achieves state-of-the-art robustness while maintaining high generation quality across a wide range of lossy scenarios.
