On the Adversarial Robustness against Natural Weather Perturbations

Yihan Wang, Yunhao Ba, Howard Chenyang Zhang, Huan Zhang, Achuta Kadambi, Stefano Soatto, Alex Wong, Cho-Jui Hsieh

Abstract:

Several algorithms are proposed to improve the robustness of deep neural networks against adversarial perturbations beyond $\ell_p$ cases, i.e. weather perturbations. However, evaluations of existing robust training algorithms are over-optimistic. This is in part due to the lack of a standardized evaluation protocol across various robust training algorithms, leading to ad-hoc methods that test robustness on either random perturbations or the adversarial samples from generative models that are used for robust training, which is either uninformative of the worst case, or is heavily biased.
In this paper, we identify such evaluation bias in these existing works and propose the first standardized and fair evaluation that compares various robust training algorithms by using physics simulators for common adverse weather effects i.e. rain and snow. Additionally, our framework identified the lack of diversity in existing robust training algorithms. As a step to address this, we propose a light-weight generative adversarial network (GAN) with improved diverse weather effects controlled by latent codes that can be used in robust training.
The proposed robust training algorithm is evaluated on two streetview classification datasets (BIC\_GSV, Places365), where it outperforms other robust training approaches based on generative models for worst-case adversarial rain and snow attacks.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Social Aspects of Machine Learning (eg, AI safety, fairness, privacy, interpretability, human-AI interaction, ethics)