Transferable Perceptual-constrained Adversarial Meshes (TPAM) is a method for generating adversarial meshes that disrupt the predictions of mesh classification networks, ensuring stealthiness, and possessing transferability to various mesh classifiers.
- Clone this repository:
git clone https://github.com/Tengjia-Kang/TPAM.git
cd TPAM- Create a virtual environment.
conda create -n TPAM python=3.8
conda activate TPAM
pip install -r requirements.txtOur attack is in a black setting, it could be divided into two part in our pipeline.
Firstly, training a surrogate network for the target victim classifier.
Secondly, attack surrogate network to obtain adversarial meshes that could transfer attack target models.
The datasets we use are consistent with the datasets used in the mesh classification we aim to attack.
To get the raw datasets go to the relevant website.
MeshCNN MeshNet PD-MeshNet MeshWalker RIMeshGNN SubdivNet ExMeshCNN
Organize the dataset with the predict logits from the target classifier.
python data/dataset_prepare.py shrec11The predictions obtained by querying the target network. For different attack targets, you may need to implement different scripts, you can refer to the fork version in my repositories.
train a surrogate network for target.
python surrogate_train/imitating_network_train.pypython attack/attack_mesh.pyIf you have any questions or issues running this code, please open an issue so we can know to fix it, or send an email to author.
This code design was adopted from Random-Walks-for-Adversarial-Meshes Public.
If you find our TPAM method is useful to your research, please cite it as follows:
@inproceedings{10.2312:pg.20241285,
booktitle = {Pacific Graphics Conference Papers and Posters},
editor = {Chen, Renjie and Ritschel, Tobias and Whiting, Emily},
title = {{TPAM: Transferable Perceptual-constrained Adversarial Meshes}},
author = {Kang, Tengjia and Li, Yuezun and Zhou, Jiaran and Xin, Shiqing and Dong, Junyu and Tu, Changhe},
year = {2024},
publisher = {The Eurographics Association},
ISBN = {978-3-03868-250-9},
DOI = {10.2312/pg.20241285}
}