Skip to content

Official implementation of the paper "Increasing Confidence in Adversarial Robustness Evaluations"

Notifications You must be signed in to change notification settings

google-research/active-adversarial-tests

Repository files navigation

Increasing Confidence in Adversarial Robustness Evaluations

This is the official repository of the paper Increasing Confidence in Adversarial Robustness Evaluations by Zimmermann et al. 2022.

The reference implementation of our proposed active test is in active_tests/decision_boundary_binarization.py, and the code to reproduce our experimental findings is in case_studies. Note, that when evaluating the defense of our authors we always used their reference implementation and only performed the minimal modification to integrate our test in their respective code base.

Citing

If you use this library, you can cite our paper. Here is an example BibTeX entry:

@inproceedings{zimmermann2022increasing,
    title={Increasing Confidence in Adversarial Robustness Evaluations},
    author={Roland S. Zimmermann and Wieland Brendel and Florian Tramer and Nicholas Carlini},
    booktitle={Advances in Neural Information Processing Systems},
    editor={Alice H. Oh and Alekh Agarwal and Danielle Belgrave and Kyunghyun Cho},
    year={2022},
    url={https://openreview.net/forum?id=NkK4i91VWp}
}

Disclaimer: This is not an official Google product.