The article discusses the challenges of protecting image copyright in the context of diffusion models (DMs), which are advanced image synthesis methods. The authors propose a novel technique, Robust Invisible Watermarking (RIW), that uses adversarial example techniques to embed invisible watermarks in images. This method ensures a high extraction accuracy of 96% for the invisible watermark after editing, significantly superior to conventional methods. The technique could be a valuable tool for asserting image ownership and protecting against unauthorized image use and deceptive edits.
Publication date: 27 Nov 2023
Project Page: https://github.com/BennyTMT/RIW
Paper: https://arxiv.org/pdf/2311.13713
