Learn2Perturb: a noise injection method for adversarial robustness

(Pytorch 1.0)

This repository contains an implementation corresponding to our CVPR 2020 paper: "Learn2Perturb: an End-to-end Feature Perturbation Learning to Improve Adversarial Robustness". A brief presentation of our work is available at this youtube link.

If you find our work useful, please cite it as follows:

@inproceedings{jeddi2020learn2perturb,
  title={Learn2Perturb: an End-to-end Feature Perturbation Learning to Improve Adversarial Robustness},
  author={Jeddi, Ahmadreza and Shafiee, Mohammad Javad and Karg, Michelle and Scharfenberger, Christian and Wong, Alexander},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={1241-1250},
  year={2020}
}

This repository includes PyTorch implementation of:

• Adversarial attacks
  • FGSM
  • PGD
  • EOT (Expectation Over Transformations [1])
• Baseline models used in experiments
• Learn2Perturb Modules

Pytorch implementations for other adversarial attacks used in this work: C&W and few-pixel attack

References

• [1] Athalye, A., Engstrom, L., Ilyas, A., and Kwok, K. Syn- thesizing robust adversarial examples. arXiv preprint arXiv:1707.07397, 2017.