The Stable Signature: Rooting Watermarks in Latent Diffusion Models
This addresses ethical concerns for users and developers of generative AI by enabling traceability of generated images, though it is an incremental improvement combining existing watermarking with latent diffusion models.
The paper tackles the problem of responsible deployment of generative image models by introducing Stable Signature, a method that embeds invisible watermarks in all generated images for detection and identification, achieving over 90% accuracy in detecting origin even after significant modifications like cropping to 10% of content.
Generative image modeling enables a wide range of applications but raises ethical concerns about responsible deployment. This paper introduces an active strategy combining image watermarking and Latent Diffusion Models. The goal is for all generated images to conceal an invisible watermark allowing for future detection and/or identification. The method quickly fine-tunes the latent decoder of the image generator, conditioned on a binary signature. A pre-trained watermark extractor recovers the hidden signature from any generated image and a statistical test then determines whether it comes from the generative model. We evaluate the invisibility and robustness of the watermarks on a variety of generation tasks, showing that Stable Signature works even after the images are modified. For instance, it detects the origin of an image generated from a text prompt, then cropped to keep $10\%$ of the content, with $90$+$\%$ accuracy at a false positive rate below 10$^{-6}$.