Research map / Adversarial robustness and security / Black-box and transfer attacks
Model Watermarking for Image Processing Networks
Jie Zhang, Dongdong Chen, Jing Liao, Han Fang, Weiming Zhang, Wenbo Zhou, Hao Cui, Nenghai Yu
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 |
|---|---|---|---|
| Use deep invisible image watermarking instead of traditional spatial bit watermarking | Image-processing model watermarking · Debone and derain, surrogate models with different architectures | success rate: 0% -> 100% | Improved |
| Add adversarial training for the watermark extractor (ablation) | Deep invisible watermarking framework · Debone and derain, UNet + L2 surrogate | success rate: 0% -> 100% | Improved |
| Add a clean loss that trains the extractor to return blank on watermark-free images (ablation) | Watermark extractor · Clean images from domains A and B | false watermark extraction | Improved |
| Add a consistent loss across extracted watermarks (ablation) | Watermark extractor · Watermarked outputs including surrogate outputs | watermark extraction quality | Improved |
| Combine task-specific image processing with watermark embedding in H | Debone model · Debone | PSNR: 25.81 -> 24.49 | Hurt |
| Combine task-specific image processing with watermark embedding in H | Debone model · Debone | SSIM: 0.91 -> 0.91 | No clear effect |
| Remove adversarial training (ablation) | Deep invisible watermarking framework · Surrogate models trained with L2 loss | success rate: 100% -> 0% | Hurt |
Closest papers by meaning
- Deep Model Intellectual Property Protection via Deep Watermarking (TPAMI 2021 · similarity 0.95)
- Robust Model Watermarking for Image Processing Networks via Structure Consistency (TPAMI 2024 · similarity 0.91)
- Lightweight-Mark: Rethinking Deep Learning-Based Watermarking (ICML 2025 · similarity 0.87)
- Safe and Robust Watermark Injection with a Single OoD Image (ICLR 2024 · similarity 0.87)
- Robustness of AI-Image Detectors: Fundamental Limits and Practical Attacks (ICLR 2024 · similarity 0.86)
- Defense against Model Extraction Attack by Bayesian Active Watermarking (ICML 2024 · similarity 0.85)
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.
