Accepted by ACM Web Conference 2025 (WWW Companion 2025), Sydney, Australia

Saliency-Aware Diffusion Reconstruction for Effective Invisible Watermark Removal[Paper]

Inzamamul Aalm, Md Tanvir Islam, Simon S. Woo^*

| Department of Computer Science and Engineering, Sungkyunkwan University, South Korea | *Corresponding Author |

---

Dependencies

pip install -r requirements.txt

We use CUDA 11.3 for our experiment.

Run the attack

python run.py

Cite this Paper

If you find our work useful in your research, please consider citing our paper and star ✨✨ this repository. Thank you!

@inproceedings{alam2025saliency,
author = {Alam, Inzamamul and Islam, Md Tanvir and Woo, Simon S.},
title = {Saliency-Aware Diffusion Reconstruction for Effective Invisible Watermark Removal},
publisher = {Association for Computing Machinery},
doi = {10.1145/3701716.3715519},
booktitle = {Companion Proceedings of the ACM on Web Conference 2025},
keywords = {adversarial attack, generative adversarial attack, regeneration attack, watermark elimination, watermark removal},
location = {Sydney NSW, Australia},
series = {WWW '25},
pages = {849–853},
numpages = {5},
year = {2025}
}

Part of this project is based on WatermarkAttacker.