Implementations of Backdoor Attacks

Prerequisites

We need the following:

• conda or miniconda (preferred)
• GPU or CPU

Setup the environment

Clone the repository. The setup script to initialize and activate the environment is collected in etc/setup_env. Simply run the following command:

. etc/setup_env

Repository artifacts

• python: code folder
• requirements.txt: list of python reqs
• README.md: this doc, and light documentation of this repos.

Implementations

Marksman Backdoor: Backdoor Attacks with Arbitrary Target Class (NeurIPS2022).

• Paper

Please refer to marksman_conditional_trigger_generation.py and marksman_conditional_backdoor_injection.py

LIRA: Learnable, Imperceptible and Robust Backdoor Attacks (ICCV2021).

• Paper
• Stage 1: Trigger Generation - LIRA learns to generate the trigger in Stage 1. Examples:
  • MNIST

    . etc/setup_env
    nohup python python/lira_trigger_generation.py --dataset mnist --clsmodel mnist_cnn --path experiments/ --epochs 10  --train-epoch 1 --mode all2one --target_label 0 --epochs_per_external_eval 10 --cls_test_epochs 5 --verbose 2 --batch-size 128 --alpha 0.5 --eps 0.1 --avoid_cls_reinit 2>&1 >experiments/logs/mnist_trigger_generation.log &
    

  • CIFAR10

    . etc/setup_env
    nohup python python/lira_trigger_generation.py --dataset cifar10 --clsmodel vgg11 --path experiments/ --epochs 50 --train-epoch 1 --mode all2one --target_label 0 --epochs_per_external_eval 10 --cls_test_epochs 5 --verbose 2 --batch-size 128 --alpha 0.5 --eps 0.1 --avoid_cls_reinit 2>&1 >experiments/logs/cifar10_trigger_generation.log &	
    

• Stage 2: Backdoor Injection. After the trigger is learned, LIRA poison and fine-tune the classifier in Stage 2. Examples:
  • MNIST

    . etc/setup_env
    nohup python python/lira_backdoor_injection.py --dataset mnist --clsmodel mnist_cnn --path experiments/ --epochs 50 --train-epoch 1 --mode all2one --target_label 0 --epochs_per_external_eval 10 --cls_test_epochs 5 --verbose 2 --batch-size 128 --alpha 0.5 --eps 0.1 --avoid_cls_reinit --test_eps 0.01 --test_alpha 0.5 --test_epochs 50 --test_lr 0.01 --schedulerC_lambda 0.1 --schedulerC_milestones 10,20,30,40 2>&1 >experiments/logs/mnist_backdoor_injection.log &	
    

  • CIFAR10

    . etc/setup_env
    nohup python python/lira_backdoor_injection.py --dataset cifar10 --clsmodel vgg11 --path experiments/ --epochs 50 --train-epoch 1 --mode all2one --target_label 0 --epochs_per_external_eval 10 --cls_test_epochs 5 --verbose 2 --batch-size 128 --alpha 0.5 --eps 0.1 --avoid_cls_reinit --test_eps 0.01 --test_alpha 0.5 --test_epochs 500 --test_lr 0.01 --schedulerC_lambda 0.1 --schedulerC_milestones 100,200,300,400 2>&1 >experiments/logs/cifar10_backdoor_injection.log &		
    

Please cite the paper, as below, when using this repository:

@inproceedings{doan2021lira,
  title={Lira: Learnable, imperceptible and robust backdoor attacks},
  author={Doan, Khoa and Lao, Yingjie and Zhao, Weijie and Li, Ping},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={11966--11976},
  year={2021}
}