Random Normalization Aggregation for Adversarial Defense

Environment

• torch 1.7.1
• torchvision 0.8.2
• torchattacks 3.2.6

Training of RNA

• To train a ResNet18 with RNA on CIFAR-10:

python train.py --random_norm_training --mixed --network ResNet18 --batch_size 128 --num_group_schedule 0 0 --worker 4 --random_type bn --gn_type gnr --save_dir resnet_c10_RNA

Evaluation of RNA

• To evaluate the performance of ResNet18 with RNA on CIFAR-10:

python evaluate.py --random_norm_training --mixed --network ResNet18 --attack_type pgd --batch_size 128 --num_group_schedule 0 0 --worker 4 --random_type bn --gn_type gnr --pretrain ./ckpt/resnet_c10.pth --save_dir resnet_c10_RNA

Pretrained models

Pretrained models are provided in google-drive, including ResNet18/WideResNet32 on CIFAR-10/100 trained by RNA.

Acknowledgement

Our codes are modified from Double-Win Quant.