Website • Paper • Doc • Leaderboard
BackdoorBench is a comprehensive benchmark of backdoor learning, which studies the adversarial vulnerablity of deep learning models in the training stage. It aims to provide easy implementations of mainstream backdoor attack and defense methods.
We disclose the backdoor model we used and the corresponding backdoor attack image in the link below. Each zip file contains the following things:
- bd_train_dataset: train backdoor data
- bd_test_dataset: test backdoor data
- attack_result.py: the backdoor model and the module that reads data
- cross_test_dataset: cross mode data during training(for some special backdoor attack: wanet, inputaware and so on)
If you want to use the backdoor model, you can download the zip file and unzip in your own workspace. Then you can use the function load_attack_result in the file save_load_attack.py to load the backdoor model, the poisoned train data and the poisoned test data.
✅ Since the previously shared document was inaccessible, we have now re-shared the backdoor model, and the link has been updated for your convenience. You can directly download the model using the link provided above.
✅ Methods:
❗ For V2.0 please check here
❗ For V1.0 please check here
BackdoorBench has the following features:
⭐️ Methods:
- 16 Backdoor attack methods: BadNets, Blended, Blind, BppAttack, CTRL, FTrojan, Input-aware, LC, LF, LIRA, PoisonInk, ReFool, SIG, SSBA, TrojanNN, WaNet
- 28 Backdoor defense/detection methods: ABL, AC, ANP, CLP, D-BR, D-ST, DBD, EP, BNP, FP, FT, FT-SAM, I-BAU, MCR, NAB, NAD, NC, NPD, RNP, SAU, SS, STRIP, BEATRIX, SCAN, SPECTRE, SS, AGPD, SentiNet, STRIP, TeCo
⭐️ Datasets: CIFAR-10, CIFAR-100, GTSRB, Tiny ImageNet
⭐️ Models: PreAct-Resnet18, VGG19_bn, ConvNeXT_tiny, ViT_B_16, VGG19, DenseNet-161, MobileNetV3-Large, EfficientNet-B3
⭐️ Learboard: We provide a public leaderboard of evaluating all backdoor attacks against all defense methods.
BackdoorBench will be continuously updated to track the lastest advances of backddor learning. The implementations of more backdoor methods, as well as their evaluations are on the way. You are welcome to contribute your backdoor methods to BackdoorBench.
You can run the following script to configure the necessary environment.
git clone git@github.com:SCLBD/BackdoorBench.git
cd BackdoorBench
conda create -n backdoorbench python=3.8
conda activate backdoorbench
sh ./sh/install.sh
sh ./sh/init_folders.sh
You can also download backdoorbench by pip.
pip install -i https://test.pypi.org/simple/ backdoorbench
The pip version of backdoorbench can be viewed at this link
This is a example for BadNets
- Generate trigger
If you want to change the trigger for BadNets, you should go to the ./resource/badnet, and follow the readme there to generate new trigger pattern.
python ./resource/badnet/generate_white_square.py --image_size 32 --square_size 3 --distance_to_right 0 --distance_to_bottom 0 --output_path ./resource/badnet/trigger_image.pngNote that for data-poisoning-based attacks (BadNets, Blended, Label Consistent, Low Frequency, SSBA).
Our scripts in ./attack are just for training, they do not include the data generation process.(Because they are time-comsuming, and we do not want to waste your time.)
You should go to the ./resource folder to generate the trigger for training.
- Backdoor training
python ./attack/badnet.py --yaml_path ../config/attack/prototype/cifar10.yaml --patch_mask_path ../resource/badnet/trigger_image.png --save_folder_name badnet_0_1
After attack you will get a folder with all files saved in ./record/<folder name in record>, including attack_result.pt for attack model and backdoored data, which will be used by following defense methods.
If you want to change the args, you can both specify them in command line and in corresponding YAML config file (eg. default.yaml).(They are the defaults we used if no args are specified in command line.)
The detailed descriptions for each attack may be put into the add_args function in each script.
Note that for some attacks, they may need pretrained models to generate backdoored data. For your ease, we provide various data/trigger/models we generated in google drive. You can download them at here (including clean_model files, ssba,lf,lc generated triggers/samples for you convenience.)
This is a demo script of running abl defense on cifar-10 for badnet attack. Before defense you need to run badnet attack on cifar-10 at first. Then you use the folder name as result_file.
python ./defense/abl.py --result_file badnet_0_1 --yaml_path ./config/defense/abl/cifar10.yaml --dataset cifar10
If you want to change the args, you can both specify them in command line and in corresponding YAML config file (eg. default.yaml).(They are the defaults we used if no args are specified in command line.)
The detailed descriptions for each attack may be put into the add_args function in each script.
| File name | Paper | |
|---|---|---|
| STRIP | strip.py | STRIP: A Defence Against Trojan Attacks on Deep Neural Networks ACSAC 2019 |
| BEATRIX | beatrix.py | The Beatrix Resurrections: Robust Backdoor Detection via Gram Matrices NDSS 2023 |
| SCAN | scan.py | Demon in the Variant: Statistical Analysis of DNNs for Robust Backdoor Contamination Detection USENIX Security 21 |
| SPECTRE | spectre.py | SPECTRE: Defending Against Backdoor Attacks Using Robust Statistics ICML 2020 |
| SS | spectral.py | Spectral Signatures in Backdoor Attacks NeurIPS 2018 |
| AGPD | agpd.py | Activation Gradient based Poisoned Sample Detection Against Backdoor Attacks arXiv |
| File name | Paper | |
|---|---|---|
| SentiNet | sentinet.py | Sentinet: Detecting localized universal attacks against deep learning systems 2020 IEEE Security and Privacy Workshops (SPW) |
| STRIP | strip.py | STRIP: A Defence Against Trojan Attacks on Deep Neural Networks ACSAC 2019 |
| TeCo | teco.py | Detecting Backdoors During the Inference Stage Based on Corruption Robustness Consistency CVPR 2023 |
We disclose the backdoor model we used and the corresponding backdoor attack image in the link below. Each zip file contains the following things:
- bd_train_dataset: train backdoor data
- bd_test_dataset: test backdoor data
- attack_result.py: the backdoor model and the module that reads data
- cross_test_dataset: cross mode data during training(for some special backdoor attack: wanet, inputaware and so on)
If you want to use the backdoor model, you can download the zip file and unzip in your own workspace. Then you can use the function load_attack_result in the file save_load_attack.py to load the backdoor model, the poisoned train data and the poisoned test data.
We recommend you to download the data and model from model zoo of our website , but we also provide the whole onedrive link at Backdoor Model as an alternative source for your convenience.
| File name | Method | Category |
|---|---|---|
| visual_tsne.py | T-SNE, the T-SNE of features | Data Analysis |
| visual_umap.py | UMAP, the UMAP of features | Data Analysis |
| visual_quality.py | Image Quality, evaluating the given results using some image quality metrics | Data Analysis |
| visual_na.py | Neuron Activation, the activation value of a given layer of Neurons | Model Analysis |
| visual_shap.py | Shapely Value, the Shapely Value for given inputs and a given layer | Model Analysis |
| visual_gradcam.py | Grad-CAM, the Grad-CAM for given inputs and a given layer | Model Analysis |
| visualize_fre.py | Frequency Map, the Frequency Saliency Map for given inputs and a given layer | Model Analysis |
| visual_act.py | Activated Image, the top images who activate the given layer of Neurons most | Model Analysis |
| visual_fv.py | Feature Visualization, the synthetic images which activate the given Neurons | Model Analysis |
| visual_fm.py | Feature Map, the output of a given layer of CNNs for a given image | Model Analysis |
| visual_actdist.py | Activation Distribution, the class distribution of Top-k images which activate the Neuron most | Model Analysis |
| visual_tac.py | Trigger Activation Change, the average (absolute) activation change between images with and without triggers | Model Analysis |
| visual_lips.py | Lipschitz Constant, the lipschitz constant of each neuron | Model Analysis |
| visual_landscape.py | Loss Landscape, the loss landscape of given results with two random directions | Model Analysis |
| visual_network.py | Network Structure, the Network Structure of given model | Model Analysis |
| visual_hessian.py | Eigenvalues of Hessian, the dense plot of hessian matrix for a batch of data | Model Analysis |
| visual_metric.py | Metrics, evaluating the given results using some metrics | Evaluation |
| visual_cm.py | Confusion Matrix |
If interested, you can read our recent works about backdoor learning, and more works about trustworthy AI can be found here.
@inproceedings{backdoorbench,
title={BackdoorBench: A Comprehensive Benchmark of Backdoor Learning},
author={Wu, Baoyuan and Chen, Hongrui and Zhang, Mingda and Zhu, Zihao and Wei, Shaokui and Yuan, Danni and Shen, Chao},
booktitle={Thirty-sixth Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
year={2022}
}
@article{wu2023adversarial,
title={Adversarial Machine Learning: A Systematic Survey of Backdoor Attack, Weight Attack and Adversarial Example},
author={Wu, Baoyuan and Liu, Li and Zhu, Zihao and Liu, Qingshan and He, Zhaofeng and Lyu, Siwei},
journal={arXiv preprint arXiv:2302.09457},
year={2023}
}
@article{wu2023defenses,
title={Defenses in Adversarial Machine Learning: A Survey},
author={Wu, Baoyuan and Wei, Shaokui and Zhu, Mingli and Zheng, Meixi and Zhu, Zihao and Zhang, Mingda and Chen, Hongrui and Yuan, Danni and Liu, Li and Liu, Qingshan},
journal={arXiv preprint arXiv:2312.08890},
year={2023}
}
@article{cheng2023tat,
title={TAT: Targeted backdoor attacks against visual object tracking},
author={Cheng, Ziyi and Wu, Baoyuan and Zhang, Zhenya and Zhao, Jianjun},
journal={Pattern Recognition},
volume={142},
pages={109629},
year={2023},
publisher={Elsevier}
}
@inproceedings{sensitivity-backdoor-defense-nips2022,
title = {Effective Backdoor Defense by Exploiting Sensitivity of Poisoned Samples},
author = {Chen, Weixin and Wu, Baoyuan and Wang, Haoqian},
booktitle = {Advances in Neural Information Processing Systems},
volume = {35},
pages = {9727--9737},
year = {2022}
}
@inproceedings{dbd-backdoor-defense-iclr2022,
title={Backdoor Defense via Decoupling the Training Process},
author={Huang, Kunzhe and Li, Yiming and Wu, Baoyuan and Qin, Zhan and Ren, Kui},
booktitle={International Conference on Learning Representations},
year={2022}
}
@inproceedings{ssba-backdoor-attack-iccv2021,
title={Invisible backdoor attack with sample-specific triggers},
author={Li, Yuezun and Li, Yiming and Wu, Baoyuan and Li, Longkang and He, Ran and Lyu, Siwei},
booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
pages={16463--16472},
year={2021}
}
@article{zhu2023vdc,
title={VDC: Versatile Data Cleanser for Detecting Dirty Samples via Visual-Linguistic Inconsistency},
author={Zhu, Zihao and Zhang, Mingda and Wei, Shaokui and Wu, Bingzhe and Wu, Baoyuan},
journal={arXiv preprint arXiv:2309.16211},
year={2023}
}
@article{zhu2023boosting,
title={Boosting backdoor attack with a learnable poisoning sample selection strategy},
author={Zhu, Zihao and Zhang, Mingda and Wei, Shaokui and Shen, Li and Fan, Yanbo and Wu, Baoyuan},
journal={arXiv preprint arXiv:2307.07328},
year={2023}
}
@article{wang2023robust,
title={Robust Backdoor Attack with Visible, Semantic, Sample-Specific, and Compatible Triggers},
author={Wang, Ruotong and Chen, Hongrui and Zhu, Zihao and Liu, Li and Zhang, Yong and Fan, Yanbo and Wu, Baoyuan},
journal={arXiv preprint arXiv:2306.00816},
year={2023}
}
@article{yuan2023activation,
title={Activation Gradient based Poisoned Sample Detection Against Backdoor Attacks},
author={Yuan, Danni and Wei, Shaokui and Zhang, Mingda and Liu, Li and Wu, Baoyuan},
journal={arXiv preprint arXiv:2312.06230},
year={2023}
}
@article{wei2023shared,
title={Shared adversarial unlearning: Backdoor mitigation by unlearning shared adversarial examples},
author={Wei, Shaokui and Zhang, Mingda and Zha, Hongyuan and Wu, Baoyuan},
journal={arXiv preprint arXiv:2307.10562},
year={2023}
}
@inproceedings{
zhu2023neural,
title={Neural Polarizer: A Lightweight and Effective Backdoor Defense via Purifying Poisoned Features},
author={Mingli Zhu and Shaokui Wei and Hongyuan Zha and Baoyuan Wu},
booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
year={2023},
url={https://openreview.net/forum?id=VFhN15Vlkj}
}
@InProceedings{Zhu_2023_ICCV,
author = {Zhu, Mingli and Wei, Shaokui and Shen, Li and Fan, Yanbo and Wu, Baoyuan},
title = {Enhancing Fine-Tuning Based Backdoor Defense with Sharpness-Aware Minimization},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
month = {October},
year = {2023},
pages = {4466-4477}
}
@article{liang2023badclip,
title={BadCLIP: Dual-Embedding Guided Backdoor Attack on Multimodal Contrastive Learning},
author={Liang, Siyuan and Zhu, Mingli and Liu, Aishan and Wu, Baoyuan and Cao, Xiaochun and Chang, Ee-Chien},
journal={arXiv preprint arXiv:2311.12075},
year={2023}
}
This repository is licensed by The Chinese University of Hong Kong, Shenzhen under Creative Commons Attribution-NonCommercial 4.0 International Public License (identified as CC BY-NC-4.0 in SPDX). More details about the license could be found in LICENSE.
This project is built by the Longgang District Key Laboratory of Intelligent Digital Economy Security (iDES) at The Chinese University of Hong Kong, Shenzhen, directed by Professor Baoyuan Wu. iDES focuses on research of trustworthy AI, including backdoor learning, adversarial examples, federated learning, fairness, etc.
If any suggestion or comment, please contact us at wubaoyuan@cuhk.edu.cn.