[IJCAI 2024] DarkFed: A Data-Free Backdoor Attack in Federated Learning

This repository comprises of implementation of DarkFed.

Installation

This code follows the setting in 3DFed(https://github.com/haoyangliASTAPLE/3DFed) withpython=3.6.13, torch=1.7.0 and torchvision=0.8.1.

• Install all dependencies using the requirements.txt in utils folder: pip install -r utils/requirements.txt.
• Install PyTorch for your CUDA version and install hdbscan~=0.8.15.
• Download the pretrained model (https://drive.google.com/file/d/11-axzUN-PTbeJCkx2KLZ2cJt15Z0fbLE/view?usp=sharing) and put it in the directory saved_models/resume_model.

Experiments on CIFAR-10

python DataFreeTraining.py --name cifar --params configs/cifar_fed.yaml

YAML files configs/cifar_fed.yaml stores the configuration for experiments.

Citation

Please cite with the below bibTex if you find it helpful to your research.

@article{DarkFed,
  title={DarkFed: A Data-Free Backdoor Attack in Federated Learning},
  author={Li, Minghui and Wan, Wei and Ning, Yuxuan and Hu, Shengshan and Xue, Lulu and Zhang, Leo Yu and Wang, Yichen},
  journal={arXiv preprint arXiv:2405.03299},
  year={2024}
}