X-Transfer Attacks: Towards Super Transferable Adversarial Attacks on CLIP

Baseline attacker GD-UAP used in ICML2025 paper "X-Transfer Attacks: Towards Super Transferable Adversarial Attacks on CLIP"

Code: https://github.com/HanxunH/XTransferBench

X-TransferBench

X-TransferBench is an open-source benchmark that provides a comprehensive collection of UAPs/TUAPs capable of achieving universal adversarial transferability. These UAPs can simultaneously transfer across data, domains, models, and tasks. Essentially, they represent perturbations that can transform any sample into an adversarial example, effective against any model and for any task.

Model Details

Surrogate Model: ResNet
Surrogate Dataset:
Threat Model: L_inf_eps=12/255
Perturbation Size: 3 x 513 x 513

Model Usage

from XTransferBench import attacker

attacker = XTransferBench.zoo.load_attacker("linf_non_targeted", "gd_uap_dl_resnet_msc_with_all_data")
images = # torch.Tensor [b, 3, h, w], values should be between 0 and 1
adv_images = attacker(images) # adversarial examples

Citation

If you use this model in your work, please cite the accompanying paper:

@article{mopuri2018generalizable,
  title={Generalizable data-free objective for crafting universal adversarial perturbations},
  author={Mopuri, Konda Reddy and Ganeshan, Aditya and Babu, R Venkatesh},
  journal={TPAMI},
  year={2018},
}

@inproceedings{
huang2025xtransfer,
title={X-Transfer Attacks: Towards Super Transferable Adversarial Attacks on CLIP},
author={Hanxun Huang and Sarah Erfani and Yige Li and Xingjun Ma and James Bailey},
booktitle={ICML},
year={2025},
}

Security and Ethical Use Statement

The perturbations provided in this project are intended solely for research purposes. They are shared with the academic and research community to advance understanding of super transferable attacks and defenses.