Focused-Flip-Federated-Backdoor-Attack

This is the official code for our paper On the Vulnerability of Backdoor Defenses for Federated Learning (accepted by AAAI-2023) by Pei Fang and Jinghui Chen .

Prerequisites

• Python (3.8+, is a must)
• Pytorch (1.11)
• CUDA (1.10+)
• some other packages (just conda install or pip install)

Step to run

1. get into the directory Focused-Flip-Federated-Backdoor-Attack/

2. get Tiny-ImageNet dataset

   wget http://cs231n.stanford.edu/tiny-imagenet-200.zip
   unzip tiny-imagenet-200.zip

3. run Bases.py

   python Bases.py 
   --defense {fedavg,ensemble-distillation,mediod-distillation,fine-tuning,mitigation-pruning,robustlr,certified-robustness,bulyan,deep-sight} 
   --config {cifar,imagenet} 
   --backdoor {ff,dba,naive,neurotoxin}
   --model {simple,resnet18}

   Hyperparameters about attack and defense baselines are mostly in Params.py, hyperparameters about dataset are mostly in configs/

Citation

Please check our paper for technical details and full results. If you find our paper useful, please cite:

@article{fang2023vulnerability,
  title={On the Vulnerability of Backdoor Defenses for Federated Learning},
  author={Fang, Pei and Chen, Jinghui},
  journal={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2023}
}