Research map / Image generation / Autoregressive and latent generation
Tree-Rings Watermarks: Invisible Fingerprints for Diffusion Images
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use ring-pattern key instead of random Gaussian key (ablation) | Tree-Ring watermark · Stable Diffusion, 75 deg rotation | AUC: 0.486 -> 0.935 | Helped |
| use ring-pattern key instead of random Gaussian key (ablation) | Tree-Ring watermark · Stable Diffusion, avg adversarial | AUC: 0.918 -> 0.975 | Helped |
| use ring-pattern key instead of zeros key (ablation) | Tree-Ring watermark · Stable Diffusion, avg adversarial | AUC: 0.963 -> 0.975 | Helped |
| use zeros key instead of ring key (ablation) | Tree-Ring watermark · Stable Diffusion, 75 deg rotation | AUC: 0.935 -> 0.994 | Helped |
| use zeros key instead of ring key (ablation) | Tree-Ring watermark · Stable Diffusion, color jitter | AUC: 0.983 -> 0.907 | Hurt |
| increase watermark radius (ablation) | Tree-RingRings · Stable Diffusion | robustness vs FID | Mixed |
| mismatch detection-time and generation-time DDIM steps (ablation) | Tree-RingRings · Stable Diffusion | AUC: 0.972 vs 0.808 range at extremes | No clear effect |
| increase guidance scale up to 18 with unknown guidance at detection (ablation) | Tree-RingRings · Stable Diffusion | AUC/TPR@1%FPR | No clear effect |
| watermark with Tree-RingRings versus no watermark | Stable Diffusion · MS-COCO FID | FID: 25.29 -> 25.93 | No clear effect |
| increase number of keys to 1000 (ablation) | Tree-RingRings · multi-key identification | identification accuracy: 0.99 -> 0.98 | No clear effect |
Closest papers by meaning
- Black-Box Forgery Attacks on Semantic Watermarks for Diffusion Models (CVPR 2025 · similarity 0.89)
- Shallow Diffuse: Robust and Invisible Watermarking through Low-Dim Subspaces in Diffusion Models (NeurIPS 2025 · similarity 0.88)
- Harnessing Frequency Spectrum Insights for Image Copyright Protection Against Diffusion Models (CVPR 2025 · similarity 0.88)
- ROBIN: Robust and Invisible Watermarks for Diffusion Models with Adversarial Optimization (NeurIPS 2024 · similarity 0.88)
- SleeperMark: Towards Robust Watermark against Fine-Tuning Text-to-image Diffusion Models (CVPR 2025 · similarity 0.88)
- T2SMark: Balancing Robustness and Diversity in Noise-as-Watermark for Diffusion Models (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.