Connecting the Digital and Physical World: Improving the Robustness of Adversarial Attacks (AAAI 2019)

This is the research code for the paper:

Steve T.K. Jan, Joseph Messou, Yen-Chen Lin, Jia-Bin Huang, Gang Wang "Connecting the Digital and Physical World: Improving the Robustness of Adversarial Attacks" In Proceedings of The Thirty-Third AAAI Conference on Artificial Intelligence (AAAI), 2019

Project page

Prerequisites

EOT.py is tested successfully in tensorflow 1.10 and python 3.6

Usage

There are several steps to generate our physical attack

1. Trained an image-to-image translation network. We used [implementations] (https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix).

• Pre-trained model.
• Data for the image-to-image network
• Here is the example script to use:

python test.py --dataroot ./datasets/test  --model test --checkpoints_dir ./checkpoints/  --name d2p/  --dataset_mode single --no_dropout  --norm batch

• ./datasets/test is the folder of images you want to transfer
• and put the model into the folder ./checkpoints/d2p/

2. Used EOT on simulated image. Here are some parameters in the top of EOT.py that can be adjusted

demo_steps=1000 # number of iterations
img_path='./simulated.JPEG' . # given an image
output_path = './demo.jpg' # output image
imagenet_json_path='./imagenet.json'
demo_epsilon = 40/255.0 # 
demo_lr = 0.5
demo_target = 21   ## adversarial class
img_class = 145  # original class

Generate noise:

python EOT.py

Citation

Please cite our paper if you find it useful for your research.

@inproceedings{aaai_d2p_2019,
  author = {Steve T.K. Jan and Joseph Messou and Yen-Chen Lin and Jia-Bin Huang and Gang Wang},
  booktitle = {The Thirty-Third AAAI Conference on Artificial Intelligence (AAAI)},
  title = {Connecting the Digital and Physical World: Improving the Robustness of
Adversarial Attacks},
  year = {2019}
}
Contact: Steve T.K. Jan (tekang at vt. edu)

Acknowledgments

EOT implemenations is referenced in (https://www.anishathalye.com/2017/07/25/synthesizing-adversarial-examples/) and thanks Yen-Chen Lin (http://yclin.me) for developing and testing