Announcing Native BM25 Ranking in AlloyDB and Cloud SQL
cloud.google.com·1d ago
AI Research
arXiv CS.AI·Jun 12·5 min read
Efficient, Robust, and Anti-Collusion Fingerprinting of Image Diffusion Models
TL;DR
Existing fingerprinting methods for text-to-image (T2I) models are vulnerable to collusion attacks, where attackers combine models to obscure fingerprints. A new method was developed that embeds user-specific identifiers into a personalized normalization module (PNM) to enhance robustness against such attacks.
✦ Why It Matters
Engineers can implement this robust fingerprinting method to protect their T2I models from unauthorized use and redistribution.
Key Takeaways
- Adopt the new fingerprinting method for T2I models.
- Utilize lossless transformations to deter collusion.
- Achieve over 99.5% accuracy in fingerprint extraction.
- Create multiple model copies without retraining.
How It Works
The proposed fingerprinting method embeds user-specific identifiers into the coefficients of a personalized normalization module (PNM) within T2I models. This allows for reliable recovery of fingerprints from generated images.
To combat collusion, the method employs lossless function-invariant parameter transformations that degrade the quality of images produced by colluded models, making them less usable.