pkuqgg / Sign-Flip-Attack

Pytorch Implementation of Sign Flip Attack(ECCV2020)

Sign-Flip-Attack

This is the Pytorch code of our ECCV2020 paper "Boosting Decision-based Black-box Adversarial Attacks with Random Sign Flip" (SFA). In this paper, we proposed a simple and efficient decision-based black-box l-inf adversarial attack.

Dependencies

• Python 3.6
• Pytorch 1.1.0
• torchvision 0.3.0
• PIL

Usage

We provide an example of how to perform targeted and untargeted attacks with SFA in test.py. original_img.png and target_img.png are randomly selected from ImageNet.
Run CUDA_VISIBLE_DEVICES=[gpu id] python test.py

Citation

If you find this work useful, please consider citing our paper. We provide a BibTeX entry of our paper below:

@inproceedings{Chen2020boosting,
    title={Boosting Decision-based Black-box Adversarial Attacks with Random Sign Flip},
    author={Chen, Weilun and Zhang, Zhaoxiang and Hu, Xiaolin and Wu, Baoyuan},
    Booktitle = {Proceedings of the European Conference on Computer Vision},
    year={2020}
}