Summary of Box-free Model Watermarks Are Prone to Black-box Removal Attacks, by Haonan An et al.
Box-Free Model Watermarks Are Prone to Black-Box Removal Attacks
by Haonan An, Guang Hua, Zhiping Lin, Yuguang Fang
First submitted to arxiv on: 16 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper reveals the vulnerabilities of box-free model watermarking in deep learning models, specifically those used for low-level image processing tasks. The technique has been shown to be effective in various aspects, but it is prone to removal attacks even when the protected model and the watermark extractor are in black boxes. To address this issue, the authors develop three removers: extractor-gradient-guided (EGG) remover, estimated gradient-based EGG remover, and transferable remover based on private proxy models. The proposed removers can successfully remove embedded watermarks while preserving image quality. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Box-free model watermarking is a technique used to protect deep learning models from being stolen or copied. This paper shows that this method has weaknesses when the protected model and the watermark extractor are not accessible. The authors create three new methods to remove these watermarks and make sure they don’t ruin the image quality. They test their methods on many images and show that they work well. |
Keywords
» Artificial intelligence » Deep learning
