SEAL: Semantic Aware Image Watermarking

Generative models have rapidly evolved to generate realistic outputs. However, their synthetic outputs increasingly challenge the clear distinction between natural and AI-generated content, necessitating robust watermarking techniques to mark synthetic images. Watermarks are typically expected to preserve the integrity of the target image, withstand removal attempts, and prevent unauthorized insertion of the watermark pattern onto unrelated images. To address this need, recent methods embed persistent watermarks into images produced by diffusion models using the initial noise of the diffusion process. Yet, to do so, they either distort the distribution of generated images or require searching a large dictionary of candidate noise patterns for detection.

In this paper, we propose a novel watermarking method that embeds semantic information about the generated image into the noise pattern, enabling a distortion-free watermark that can be verified without requiring a database of key patterns. Instead, the key pattern can be inferred from the semantic embedding of the image using locality-sensitive hashing. Furthermore, conditioning the watermark detection on the original image content improves its robustness against forgery attacks. To demonstrate that, we consider two largely overlooked attack strategies: (i) an attacker extracting the initial noise and generating a novel image with the same pattern; (ii) an attacker inserting an unrelated (potentially harmful) object into a watermarked image, while preserving the watermark. We empirically validate our method's increased robustness to these attacks. Taken together, our results suggest that content-aware watermarks can mitigate risks arising from image-generative models. Our code is available at https://github.com/Kasraarabi/ SEAL.