Exploring the Landscape of Spatial Robustness

Logan Engstrom, Brandon Tran, Dimitris Tsipras, Ludwig Schmidt, Aleksander Madry
12/7/2017
cs.LGcs.CVcs.NE

Official code: MadryLab/adversarial_spatial (47 stars, MIT, last commit 2019-09-16). 1 more implementation below.

Abstract

The study of adversarial robustness has so far largely focused on perturbations bound in 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 p-norm case, first-order methods cannot reliably find worst-case perturbations. This highlights spatial robustness as a fundamentally different setting requiring additional study. Code available at https://github.com/MadryLab/adversarial_spatial and https://github.com/MadryLab/spatial-pytorch.

View on arXivView PDF

Code Implementations(2)

MadryLab/adversarial_spatialOfficial100%
MIT

Investigating the robustness of state-of-the-art CNN architectures to simple spatial transformations.

4713PythonJul 2, 20187 years agoMIT
MadryLab/spatial-pytorchOfficial100%

Codebase for "Exploring the Landscape of Spatial Robustness" (ICML'19, https://arxiv.org/abs/1712.02779).

253Shell, PythonAug 5, 20197 years ago

Similar Papers

Robustness of Rotation-Equivariant Networks to Adversarial Perturbations

88% match1 repo752018

Attribute-Guided Adversarial Training for Robustness to Natural Perturbations

85% match1 repo32020

(Provable) Adversarial Robustness for Group Equivariant Tasks: Graphs, Point Clouds, Molecules, and More

85% match2023

Bridging Optimal Transport and Jacobian Regularization by Optimal Trajectory for Enhanced Adversarial Defense

83% match2023

Not So Robust After All: Evaluating the Robustness of Deep Neural Networks to Unseen Adversarial Attacks

83% match2023

Efficient Neural Network Robustness Certification with General Activation Functions

83% match1 repo182018

Cite this paper

@article{engstrom2017exploring,
  title  = {Exploring the Landscape of Spatial Robustness},
  author = {Logan Engstrom and Brandon Tran and Dimitris Tsipras and Ludwig Schmidt and Aleksander Madry},
  year   = {2017},
  eprint = {1712.02779},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1712.02779}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1712.02779.svg)](https://codeofpaper.com/paper/1712.02779)

Discussion