Bilevel Transfer Attack (BETAK)

This repo contains the source code for the paper named "Advancing Generalized Transfer Attack with Initialization Derived Bilevel Optimization and Dynamic Sequence Truncation", which has been accepted to IJCAI 2024.

Environment

• Python
• PyTorch
• Higher
• timm

Datasets(ImageNet)

Following BPA, we also randomly sample 5,000 images pertaining to the 1,000 categories from ILSVRC 2012 validation set, which could be classified correctly by all the victim models. The corresponding CSV file are saved as data/imagenet/selected_imagenet_resnet50.csv and data/imagenet/selected_imagenet_vgg19_bn.csv.

Before running, you should download the ILSVRC 2012 validation set and specify the directory using parameters --imagenet_val_dir. Your directory structure is as follows:

imagenet-val
    ├── n07880968
    |   ├── ILSVRC2012_val_00011685.JPEG
    |   └── ...
    ├── n02927161
    |   ├── ILSVRC2012_val_00011685.JPEG
    |   └── ...
    └── ...

Models

The pretrained models can be downloaded at here, then extract them to ckpt/.

Attack & Test with BETAK

Untargeted attack using BETAK+ BPA + PGD as an example:

python attack_eval_imagenet_betak.py --epsilon 0.03 --sgm_lambda 1.0 --niters 10 --method max_relu_silu_pgd --batch_size 25 --save_dir data/betak_imagenet/max_relu_silu_pgd_attack_resnet50_meta_incv3meta_lr_2.0_meta_step3_inner_loop10 --device_id 0 --imagenet_val_dir imagenet/val --model_name resnet50 --alpha 0.006 --inner_loop 10 --attack_lr 2.0 --meta_steps 3

You should modify the default directory for the validation set according to your system path.

Citation