Research map / Image generation / Autoregressive and latent generation
Attention to Neural Plagiarism: Diffusion Models Can Plagiarize Your Copyrighted Images!
Zihang Zou, Boqing Gong, Liqiang Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add attention-perturbation shims instead of plain regeneration (Regen) with same noisy start | Neural Plagiarism attack · MS-COCO DwtDctSvd, late start | bit accuracy: 0.64 -> 0.52 | Improved |
| add attention-perturbation shims instead of Regen | Neural Plagiarism attack · MS-COCO RivaGAN, late start | bit accuracy: 0.60 -> 0.56 | No clear effect |
| add shims with early start from inverse latent vs Regen | Neural Plagiarism attack · MS-COCO DwtDctSvd, early start | FID: 21.36 -> 71.14 | Improved |
| add shims with early start from inverse latent vs Regen | Neural Plagiarism attack · MS-COCO RivaGAN, early start | bit accuracy: 0.59 -> 0.52 | Improved |
| start from inverse latent instead of noisy latent in late-start setting (ablation) | Neural Plagiarism attack · invisible watermark removal | semantic change | Hurt |
| attack Stable Signature watermark with latent shims | Neural Plagiarism attack · MS-COCO Stable Signature | T@1%F: 1.00 -> 0.00 | Improved |
| attack Tree-Ring watermark with latent shims | Neural Plagiarism attack · MS-COCO Tree-Ring | T@1%F: 1.00 -> 0.98 | No clear effect |
| replace RivaGAN watermark with new watermark after attack | Neural Plagiarism attack · RivaGAN ambiguity attack | bit accuracy: BA(w*) 0.75 -> 1.00; BA(w) 1.00 -> 0.75 | Improved |
| embed alternative watermark channel after attack | Neural Plagiarism attack · Tree-Ring ambiguity attack | T@1%F: T@1%F(w*) 0.00 -> 1.00; T@1%F(w) 1.00 -> 0.99 | Improved |
| vary shim timestep (t=100 to t=700) on real copyrighted images (ablation) | Neural Plagiarism attack · real-world copyrighted images | semantic modification | Improved |
Closest papers by meaning
- Harnessing Frequency Spectrum Insights for Image Copyright Protection Against Diffusion Models (CVPR 2025 · similarity 0.89)
- The Stronger the Diffusion Model, the Easier the Backdoor: Data Poisoning to Induce Copyright BreachesWithout Adjusting Finetuning Pipeline (ICML 2024 · similarity 0.89)
- Disguised Copyright Infringement of Latent Diffusion Models (ICML 2024 · similarity 0.88)
- SleeperMark: Towards Robust Watermark against Fine-Tuning Text-to-image Diffusion Models (CVPR 2025 · similarity 0.88)
- CGI-DM: Digital Copyright Authentication for Diffusion Models via Contrasting Gradient Inversion (CVPR 2024 · similarity 0.87)
- Watermark-embedded Adversarial Examples for Copyright Protection against Diffusion Models (CVPR 2024 · 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.
