Attacking Transformers with Feature Diversity Adversarial Perturbation

Authors: Chenxing Gao, Hang Zhou, Junqing Yu, YuTeng Ye, Jiale Cai, Junle Wang, Wei Yang

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct experiments to test our method on Vi T-based models, CNN models, and MLP models. Furthermore, we assess the cross-task transferability of our attack method.
Researcher Affiliation Collaboration Chenxing Gao1, Hang Zhou1, Junqing Yu1, Yu Teng Ye1, Jiale Cai1, Wei Yang1* 1Huazhong University of Science and Technology, Wuhan, China Junle Wang2 2Tencent
Pseudocode Yes Algorithm 1: Feature Diversity Adversarial Perturbation on Vi Ts
Open Source Code No The paper does not include any explicit statement about releasing source code or provide a link to a code repository.
Open Datasets Yes Dataset: Similar to the settings in Dong(Dong et al. 2018), we randomly select data 1000 images from the validation set Image Net 2012 (Russakovsky et al. 2015).
Dataset Splits No The paper mentions using 1000 images from the validation set of ImageNet 2012, but it does not provide specific train/validation/test dataset splits for its own experimental setup.
Hardware Specification No The computation is completed in the HPC Platform of Huazhong University of Science and Technology.
Software Dependencies No The paper does not provide specific software names with version numbers for its dependencies.
Experiment Setup Yes Attack settings: we conduct attacks using a maximum perturbation value of ϵ = 16, the total number of attack iterations is N = 30, and the step size α = 3/255.