Research map / Adversarial robustness and security / Attacks on text and graphs
Watermark Vaccine: Adversarial Attacks to Prevent Watermark Removal
Xinwei Liu, Jian Liu, Yang Bai, Jindong Gu, Tao Chen, Xiaojun Jia, Xiaochun Cao
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 DWV perturbation to host image | WDNet · Watermark removal | PSNR_h: 38.62 -> 29.68 | Improved |
| add DWV perturbation to host image | BVMR · Watermark removal | PSNR_h: 41.96 -> 29.43 | Improved |
| add DWV perturbation to host image | SplitNet · Watermark removal | PSNR_h: 42.32 -> 34.12 | Improved |
| add IWV perturbation to host image | WDNet · Watermark removal | RMSE_w_w: 52.77 -> 28.00 | Improved |
| add IWV perturbation to host image | SplitNet · Watermark removal | RMSE_w_w: 49.67 -> 35.00 | Improved |
| add random noise instead of vaccine (ablation) | WDNet · Watermark removal | PSNR_h: 38.62 -> 38.19 | No clear effect |
| transfer vaccine from WDNet to BVMR/SplitNet target | DWV · Watermark removal | RMSE_h: 3.14 -> 8.30 on WDNet, 2.78 -> 2.80 on BVMR | No clear effect |
| increase watermark size from 60 to 100 | DWV · WDNet | RMSE_h_w: 48.87 -> 37.32 | Hurt |
| increase watermark size from 60 to 100 | IWV · WDNet | RMSE_w_w: 13.70 -> 32.87 | Hurt |
| increase watermark alpha from 0.45 to 0.65 | DWV · WDNet | RMSE_h_w: 49.44 -> 39.31 | Hurt |
| increase watermark alpha from 0.45 to 0.65 | IWV · WDNet | RMSE_w_w: 18.00 -> 21.40 | Hurt |
| stack vaccines from three networks (ablation) | DWV/IWV · WDNet | protection under perturbation budgets | Improved |
Closest papers by meaning
- Watermark-embedded Adversarial Examples for Copyright Protection against Diffusion Models (CVPR 2024 · similarity 0.85)
- Decoder Gradient Shield: Provable and High-Fidelity Prevention of Gradient-Based Box-Free Watermark Removal (CVPR 2025 · similarity 0.85)
- Safe and Robust Watermark Injection with a Single OoD Image (ICLR 2024 · similarity 0.84)
- Robustness of AI-Image Detectors: Fundamental Limits and Practical Attacks (ICLR 2024 · similarity 0.83)
- Lightweight-Mark: Rethinking Deep Learning-Based Watermarking (ICML 2025 · similarity 0.83)
- ROBIN: Robust and Invisible Watermarks for Diffusion Models with Adversarial Optimization (NeurIPS 2024 · similarity 0.82)
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.
