P2FA

[ICML 2025] This repository contains the PyTorch code for the paper: Pixel2Feature Attack (P2FA): Rethinking the Perturbed Space to Enhance Adversarial Transferability

Renpu Liu, Hao Wu, Jiawei Zhang, Xin Cheng, Xiangyang Luo, Bin Ma, Jinwei Wang

Datasets

1000 images for the NIPS 2017 adversarial competition.

directory/
├── 0.png
├── 1.png
├── ...
└── 999.jpg

normalization:

import torchvision.transforms as T

transforms = T.Compose([
    T.ToTensor(),
    T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])

Pre-trained Models

Models	Packages	Accuracy (%)
Inception-v3 (Inc-v3)	Pytorch	95.4
Inception-v4 (Inc-v4)	timm	94.8
Inception-ResNet-v2 (IncRes-v2 )	timm	97.2
ResNet-50 (Res-50)	Pytorch	91.8
ResNet-152 (Res-152)	Pytorch	93.6
VGG-16 (Vgg-16)	Pytorch	85.5
VGG-19 (Vgg-19)	Pytorch	87.5
Adv-Inception-v3 (Adv-Inc-v3)	timm	86.8
Ens-Adv-Inception-ResNet-v2 (Ens-IncRes-v2)	timm	94.5

P2FA

python main.py

Citation

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

@inproceedings{liupixel2feature,
  title={Pixel2Feature Attack (P2FA): Rethinking the Perturbed Space to Enhance Adversarial Transferability},
  author={Liu, Renpu and Wu, Hao and Zhang, Jiawei and Cheng, Xin and Luo, Xiangyang and Ma, Bin and Wang, Jinwei},
  booktitle={Forty-second International Conference on Machine Learning}
}