This repository includes the source code of the paper "Detection by Attack: Detecting Adversarial Samples by Undercover Attack". Please cite our paper when you use this program! 😍 This paper has been accepted to the conference "European Symposium on Research in Computer Security (ESORICS20)". This paper can be downloaded here.
@inproceedings{zhou2020detection,
title={Detection by attack: Detecting adversarial samples by undercover attack},
author={Zhou, Qifei and Zhang, Rong and Wu, Bo and Li, Weiping and Mo, Tong},
booktitle={European Symposium on Research in Computer Security},
pages={146--164},
year={2020},
organization={Springer}
}
The pipeline of our framework consists of two steps:
- Injecting adversarial samples to train the classification model.
- Training a simple multi-layer perceptron (MLP) classifier to judge whether the sample is adversarial.
We take MNIST and CIFAR as examples: the mnist_undercover_train.py and cifar_undercover_train.py refer to the step one; the mnist_DBA.ipynb and cifar_DBA.ipynb refer to the step two.
Please let us know if you encounter any problems.
The contact email is [email protected]