This repository includes the implementation of our paper: Generate (non-software) Bugs to Fool Classifiers.

You can find generated adversarial examples at our project page.

This repository contains two different implementations for image and audio adversarial examples.

Please look for each directories for the detailed usage.

The most of the source code in this repository is based on https://github.com/tensorflow/models and https://github.com/chrisdonahue/wavegan/, which are distributed under the Apache License v2 and MIT License, respectively.

If you use this code for your research, please cite this paper:

@article{yakura2020generate,
  title={Generate (non-software) Bugs to Fool Classifiers},
  author={Yakura, Hiromu and Akimoto, Youhei and Sakuma, Jun},
  booktitle={Proceedings of the 34th {AAAI} Conference on Artificial Intelligence},
  publisher={{AAAI} Press},
  year={2020}
}