This is the official implementation of our paper BATT: Backdoor Attack with Transformation-Based Triggers, accepted by ICASSP, 2023. This research project is developed based on Python 3 and Pytorch, created by Tong Xu and Yiming Li.
If our work or this repo is useful for your research, please cite our paper as follows:
@inproceedings{xu2023batt,
title={BATT: Backdoor Attack with Transformation-Based Triggers},
author={Xu, Tong and Li, Yiming and Jiang, Yong and Xia, Shu-Tao},
booktitle={ICASSP},
year={2023},
}
To install requirements:
pip install -r requirements.txt
Make sure the directory follows:
BATT
├── train
│ ├── batt_r
│ └── batt_t
├── network
│ ├── resnet
│ └── vgg
├── data
│ ├── cifar10
│ └── GTSRB
Make sure the directory data follows:
data
├── cifar10
├── GTSRB
│ ├── train
│ └── test
📋 Data Download Link:
data
To train the BATT-R in the paper, run these commands:
CIFAR-10:
python run_batt_r.py
Get attacked model in the dictionary 'result':
result
├── attack
│ ├── batt_r_cifar10.pth
│ └── ...
📋 Sample Infected Model Download Link:
model
Thanks to Backdoorbox for providing framework codes.