SoK: Certified Robustness for Deep Neural Networks

Recently, provable (i.e. certified) adversarial robustness training and verification methods have demonstrated their effectiveness against adversarial attacks. In contrast to empirical robustness and empirical adversarial attacks, the provable robustness verification provides rigorous lower bound of robustness for a given neural network, such that no existing or future attacks will attack further.

This repo contains the leaderboard website of state-of-the-art certified robustness achieved on common datasets.

Website: https://sokcertifiedrobustness.github.io/

Accompanying SoK paper is accepted by IEEE S&P (Oakland) 2023!

If you find this repo helpful, please consider cite our paper:

@inproceedings{li2023sok,
    title={{SoK}: Certified Robustness for Deep Neural Networks},
    author={Linyi Li and Tao Xie and Bo Li},
    booktitle={44th {IEEE} Symposium on Security and Privacy, {SP} 2023, San Francisco, CA, USA, 22-26 May 2023},
    publisher={IEEE},
    year={2023}
}

Accompanying open-source toolbox: https://github.com/AI-secure/VeriGauge

Announcing your great result?

1. To add your results in trend curves of https://sokcertifiedrobustness.github.io/leaderboard/, feel free to directly edit _data/sota_trend.yml in the repo and send a pull request.

2. To add your results in tables of https://sokcertifiedrobustness.github.io/leaderboard/, feel free to directly edit _data/board.yml in the repo and send a pull request.

---

Current maintainer: Linyi Li

Contributors:

• Linyi Li
• Zayd Hammoudeh
• Alessandro De Palma
• Zifan Wang
• Václav Voráček
• Thomas R. Altstidl

(missing? please create a PR to include your name)