Code for paper "Detecting Adversarial Data Using Perturbation Forgery".

Get Started

Datasets are CIFAR10 and ImageNet100.

Our codebase accesses the datasets from ./data/ and checkpoints from ./results/checkpoints/ by default.

├── ...
├── data
│   
├── results
│   
├── main.py
├── ...

All of the adversarial data are generated using torchattacks.

Generate noise distribution by gen_dist.py or download from here. Put distributions under 'data/dist/'.

Train Detector

python main.py
--config configs/datasets/general/DIS_CIFAR10.yml
configs/pipelines/train/DIS_train_CIFAR10.yml
--force_merge True
--preprocessor.name ImageNet

Test

python main.py
--config configs/datasets/general/DIS_CIFAR10.yml
configs/pipelines/train/DIS_test_CIFAR10.yml
--force_merge True
--preprocessor.name ImageNet

The checkpoints are coming soon...

## Dependencies
python 3.8.8, PyTorch = 1.10.0, cudatoolkit = 11.7, torchvision, tqdm, scikit-learn, mmcv, numpy, opencv-python, dlib, Pillow