Exploring Adversarial Examples and Adversarial Robustness of Convolutional Neural Networks by Mutual Information

Official implementation for

• Exploring Adversarial Examples and Adversarial Robustness of Convolutional Neural Networks by Mutual Information. (arXiv)

For any questions, contact (zhangjiebao2014@mail.ynu.edu.cn).

Requirements

1. Python
2. Pytorch
3. Torattacks >= 3.2.6
4. Torchvision
5. Pytorchcv

Preparations

• some file paths will be created manually

Estimate the mutual information in normal or adversarial training

python MI_flow_in_training.py

Estimate the mutual information while the input suffer from the information distortion

python MI_flow_in_forward.py

Citation

If you find this repo useful for your research, please consider citing the paper

@misc{https://doi.org/10.48550/arxiv.2207.05756,
  doi = {10.48550/ARXIV.2207.05756},
  url = {https://arxiv.org/abs/2207.05756},
  author = {Zhang, Jiebao and Qian, Wenhua and Nie, Rencan and Cao, Jinde and Xu, Dan},
  title = {Exploring Adversarial Examples and Adversarial Robustness of Convolutional Neural Networks by Mutual Information},
  publisher = {arXiv},
  year = {2022},
}