Attack Diffusion Models for Adversarial Purification

This is a repository for storing code for course project. The code was adapted from the official implementation of the paper Diffusion Models for Adversarial Purification.

Requirements

• 64-bit Python 3.8.
• CUDA=11.0 must be installed first.

Data and pre-trained models

ImageNet data should be put in ./dataset/imagenet_lmdb/val/ and categorized into different classes in different subfolders. convert the ImageNet dataset to LMDB.

There is no need to download CIFAR-10 separately.

For the pre-trained diffusion models, you need to first download them from the following links:

• Score SDE for CIFAR-10: (vp/cifar10_ddpmpp_deep_continuous: download link)
• Guided Diffusion for ImageNet: (256x256 diffusion unconditional: download link)

For the pre-trained classifiers, most of them do not need to be downloaded separately, except for

• resnet-50 on CIFAR-10: download link

Note that you have to put all the pretrained models in the pretrained directory.

Run experiments on CIFAR-10

AutoAttack L2

bash cifar10_L2_rand.sh bash imagenet_L2_rand.sh

AutoAttack Linf

bash imagenet_Linf_rand.sh

License

Please check the LICENSE file. This work may be used non-commercially, meaning for research or evaluation purposes only. For business inquiries, please contact [email protected].