Introduction

Our paper, titled "Physical Adversarial Attack meets Computer Vision: A Decade Survey," has been published in IEEE T-PAMI. This survey aims to summarize, evaluate, and analyze existing physical adversarial attack methods, providing insights for the development of trustworthy AI.
Paper | arXiv

The repository is dedicated to tracking the latest advances in the field of Physical Adversarial Attack (PAA). The maintainer will continue to update it.

If you find any omitted literature, please feel free to submit issues for addition. Many thanks!

Table of Contents:

• Order by Task
• Order by Time
• Other Survey

News

[2024-12-09] 📊 We have released the corresponding quantitative data and the code for computing hiPAA. Please refer to code/hiPAA.m.

[2024-11-21] ⚡ We have submitted the latest version of our survey paper, Physical Adversarial Attack meets Computer Vision: A Decade Survey. Feedback and discussions are highly welcome.

[2024-07-17] 🔥 Our paper, Physical Adversarial Attack meets Computer Vision: A Decade Survey, has been accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). We are currently working on further refining this research for its fourth version.

[2022-10-01] 😐 We have submitted our Physical-Adversarial-Attack survey on arXiv: Physical Adversarial Attack meets Computer Vision: A Decade Survey. We will continue to polish this work.

Evaluation

In our survey paper, we introduce a metric named hiPAA (Hexagonal Indicator of Physical Adversarial Attack), which encompasses six dimensions: Effectiveness, Stealthiness, Robustness, Practicability, Aesthetics, and Economics. This metric provides a systematic perspective for evaluating physical adversarial attacks.

Using hiPAA, we quantitatively evaluated existing methods. We also release the corresponding quantitative data and the code for computing hiPAA. The results are presented as radar charts. For detailed instructions, please refer to code/hiPAA.m.

Citation

If you find our paper and repository useful for your research, please kindly cite as:

@article{wei2024physical,
  title={Physical Adversarial Attack Meets Computer Vision: A Decade Survey},
  author={Wei, Hui and Tang, Hao and Jia, Xuemei and Wang, Zhixiang and Yu, Hanxun and Li, Zhubo and Satoh, Shin’ichi and Van Gool, Luc and Wang, Zheng},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2024},
  volume={46},
  number={12},
  pages={9797-9817},
  doi={10.1109/TPAMI.2024.3430860}
}