Summary of Erasablemask: a Robust and Erasable Privacy Protection Scheme Against Black-box Face Recognition Models, by Sipeng Shen et al.
ErasableMask: A Robust and Erasable Privacy Protection Scheme against Black-box Face Recognition Models
by Sipeng Shen, Yunming Zhang, Dengpan Ye, Xiuwen Shi, Long Tang, Haoran Duan, Jiacheng Deng, Ziyi Liu
First submitted to arxiv on: 22 Dec 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 proposes ErasableMask, a robust and erasable privacy protection scheme for face recognition (FR) models. Existing FR-based facial privacy schemes often suffer from weak transferability against black-box models and permanently damage identifiable information, hindering authorized operations like forensics and authentication. ErasableMask addresses these limitations by introducing a novel meta-auxiliary attack that boosts black-box transferability via stable optimization strategies. It also offers a perturbation erasion mechanism to erase semantic perturbations in protected faces without degrading image quality. A curriculum learning strategy is employed to mitigate optimization conflicts between adversarial attacks and perturbation erasion. The proposed scheme achieves state-of-the-art performance on the CelebA-HQ and FFHQ datasets, demonstrating over 72% confidence on average in commercial FR systems. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about keeping people’s faces private from face recognition (FR) models. Right now, FR models can identify people easily, but they also raise serious privacy concerns. People want to keep their identities safe without sacrificing the ability to use FR for things like verifying someone’s identity or solving crimes. The researchers propose a new way to protect people’s faces called ErasableMask. It works by using special tricks to confuse the FR models and hide the identifying information, while still allowing it to be used for authorized purposes. The new method performs well on two large datasets and has the potential to make face recognition more secure and private. |
Keywords
» Artificial intelligence » Curriculum learning » Face recognition » Optimization » Transferability
