Just open-sourced AdverTorch, a Pytorch based Toolbox for Adversarial Robustness Research

Hi folks,

We (Borealis AI, Canada) have just open-sourced AdverTorch, a pytorch based toolbox for adversarial robustness research. It is still work in progress, but contains basic functionalities on attacks, defenses, BPDA, and examples of adversarial training. You’re welcomed to try it out, and let us know for questions and suggestions.

The repo is at https://github.com/borealisai/advertorch


That’s great! Thanks for sharing!

Thanks for this, I was looking for a simple way to apply adversarial samples during training!
I know it is intended for adversarial robustness, but do you think it can help classification accuracy in a similar way that AdvProp (https://arxiv.org/pdf/1911.09665v1.pdf) does?