Validation of watermark embedding in diffusion models
Abstract
This study aims to ensure the reliability of AI models, including derived versions, by identifying the developers of diffusion models utilizing digital watermarking techniques employed for DNN copyright protection. We embedded watermarks into original models and demonstrated that these watermarks remain detectable even in derived models generated through fine - tuning. Furthermore, we showed the potential for the original watermarks to coexist with separate watermarks embedded within the derived models.
Fields of Study & Subjects
Advanced Steganography and Watermarking Techniques
Law
AI
and Intellectual Property
Adversarial Robustness in Machine Learning
Publication details
- Journal / source
- International Workshop on Advanced Imaging Technology (IWAIT)
- Pages
- 103–103
- Publication date
- 27 February 2026
- Language
- EN