Exploring the Landscape of Spatial Robustness

Logan Engstrom, Brandon Tran, Dimitris Tsipras, Ludwig Schmidt, Aleksander Madry
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:1802-1811, 2019.

Abstract

The study of adversarial robustness has so far largely focused on perturbations bound in $\ell_p$-norms. However, state-of-the-art models turn out to be also vulnerable to other, more natural classes of perturbations such as translations and rotations. In this work, we thoroughly investigate the vulnerability of neural network–based classifiers to rotations and translations. While data augmentation offers relatively small robustness, we use ideas from robust optimization and test-time input aggregation to significantly improve robustness. Finally we find that, in contrast to the $\ell_p$-norm case, first-order methods cannot reliably find worst-case perturbations. This highlights spatial robustness as a fundamentally different setting requiring additional study.

Cite this Paper

BibTeX
@InProceedings{pmlr-v97-engstrom19a, title = {Exploring the Landscape of Spatial Robustness}, author = {Engstrom, Logan and Tran, Brandon and Tsipras, Dimitris and Schmidt, Ludwig and Madry, Aleksander}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {1802--1811}, year = {2019}, editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan}, volume = {97}, series = {Proceedings of Machine Learning Research}, month = {09--15 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v97/engstrom19a/engstrom19a.pdf}, url = {https://proceedings.mlr.press/v97/engstrom19a.html}, abstract = {The study of adversarial robustness has so far largely focused on perturbations bound in $\ell_p$-norms. However, state-of-the-art models turn out to be also vulnerable to other, more natural classes of perturbations such as translations and rotations. In this work, we thoroughly investigate the vulnerability of neural network–based classifiers to rotations and translations. While data augmentation offers relatively small robustness, we use ideas from robust optimization and test-time input aggregation to significantly improve robustness. Finally we find that, in contrast to the $\ell_p$-norm case, first-order methods cannot reliably find worst-case perturbations. This highlights spatial robustness as a fundamentally different setting requiring additional study.} }
Endnote
%0 Conference Paper %T Exploring the Landscape of Spatial Robustness %A Logan Engstrom %A Brandon Tran %A Dimitris Tsipras %A Ludwig Schmidt %A Aleksander Madry %B Proceedings of the 36th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Ruslan Salakhutdinov %F pmlr-v97-engstrom19a %I PMLR %P 1802--1811 %U https://proceedings.mlr.press/v97/engstrom19a.html %V 97 %X The study of adversarial robustness has so far largely focused on perturbations bound in $\ell_p$-norms. However, state-of-the-art models turn out to be also vulnerable to other, more natural classes of perturbations such as translations and rotations. In this work, we thoroughly investigate the vulnerability of neural network–based classifiers to rotations and translations. While data augmentation offers relatively small robustness, we use ideas from robust optimization and test-time input aggregation to significantly improve robustness. Finally we find that, in contrast to the $\ell_p$-norm case, first-order methods cannot reliably find worst-case perturbations. This highlights spatial robustness as a fundamentally different setting requiring additional study.
APA
Engstrom, L., Tran, B., Tsipras, D., Schmidt, L. & Madry, A.. (2019). Exploring the Landscape of Spatial Robustness. Proceedings of the 36th International Conference on Machine Learning, in Proceedings of Machine Learning Research 97:1802-1811 Available from https://proceedings.mlr.press/v97/engstrom19a.html.

Related Material