Home

Awesome

Structure Invariant Attack

This repository contains code to reproduce results from the paper:

Structure Invariant Transformation for better Adversarial Transferability (ICCV 2023)

Xiaosen Wang, Zeliang Zhang, Jianping Zhang

The transformed images generated by various transformations.

We also include the code in the framework TransferAttack.

Requirements

Qucik Start

Prepare the data and models

You should download the data offered by Admix and place it in Input/data with label file val_rs.csv in Input.

Runing attack

Taking SIA attack for example, you can run this attack as following:

CUDA_VISIBLE_DEVICES=gpuid python main.py --model model_name  

Evaluating the attack

The generated adversarial examples would be stored in directory ./outputs. Then run the file main.py with eval to evaluate the success rate of each model used in the paper:

CUDA_VISIBLE_DEVICES=gpuid python main.py --eval

Citation

If you find the idea or code useful for your research, please consider citing our paper:

@inproceedings{wang2023structure,
     title={{Structure Invariant Transformation for better Adversarial Transferability}},
     author={Xiaosen Wang and Zeliang Zhang and Jianping Zhang},
     booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
     year={2023}
}