mlsanityData to 7 Oct 2026

Robustifying ℓ_∞ Adversarial Training to the Union of Perturbation Models

arXiv 2105.14710 ↗First posted 31 May 2021cs.LG stat.MLcited by 0

Not on the map. Why?

Abstract (arXiv)

Classical adversarial training (AT) frameworks are designed to achieve high adversarial accuracy against a single attack type, typically ℓ_∞ norm-bounded perturbations. Recent extensions in AT have focused on defending against the union of multiple perturbations but this benefit is obtained at the expense of a significant (up to 10×) increase in training complexity over single-attack ℓ_∞ AT. In this work, we expand the capabilities of widely popular single-attack ℓ_∞ AT frameworks to provide robustness to the union of (ℓ_∞, ℓ2, ℓ1) perturbations while preserving their training efficiency. Our technique, referred to as Shaped Noise Augmented Processing (SNAP), exploits a well-established byproduct of single-attack AT frameworks -- the reduction in the curvature of the decision boundary of networks. SNAP prepends a given deep net with a shaped noise augmentation layer whose distribution is learned along with network parameters using any standard single-attack AT. As a result, SNAP enhances adversarial accuracy of ResNet-18 on CIFAR-10 against the union of (ℓ_∞, ℓ2, ℓ1) perturbations by 14%-to-20% for four state-of-the-art (SOTA) single-attack ℓ_∞ AT frameworks, and, for the first time, establishes a benchmark for ResNet-50 and ResNet-101 on ImageNet.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 39 of its 49 references that are in the map. First posted on arXiv 31 May 2021.

Builds on

39 of 49 references.

References · 39

  1. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  2. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,857
  3. 2019Theoretically Principled Trade-off between Robustness and Accuracycited by 1,510
  4. 2019Certified Adversarial Robustness via Randomized Smoothingcited by 1,324
  5. 2015DeepFool: a simple and accurate method to fool deep neural networkscited by 1,315
  6. 2019Adversarial Examples Are Not Bugs, They Are Featurescited by 950
  7. 2020Fast is better than free: Revisiting adversarial trainingcited by 723
  8. 2019Square Attack: a query-efficient black-box adversarial attack via random searchcited by 682
  9. 2019Adversarial Training for Free!cited by 638
  10. 2020On Adaptive Attacks to Adversarial Example Defensescited by 547
  11. 2020Overfitting in adversarially robust deep learningcited by 504
  12. 2017Decision-Based Adversarial Attacks: Reliable Attacks Against Black-Box Machine Learning Modelscited by 459
27 more
  1. 2019Provably Robust Deep Learning via Adversarially Trained Smoothed Classifierscited by 328
  2. 2020Uncovering the Limits of Adversarial Training against Norm-Bounded Adversarial Examplescited by 275
  3. 2020Attacks Which Do Not Kill Training Make Adversarial Learning Strongercited by 242
  4. 2021Fixing Data Augmentation to Improve Adversarial Robustnesscited by 201
  5. 2018Spatially Transformed Adversarial Examplescited by 195
  6. 2019Adversarial Training and Robustness for Multiple Perturbationscited by 182
  7. 2019You Only Propagate Once: Accelerating Adversarial Training via Maximal Principlecited by 181
  8. 2018Robustness via curvature regularization, and vice versacited by 174
  9. 2017EAD: Elastic-Net Attacks to Deep Neural Networks via Adversarial Examplescited by 154
  10. 2020Randomized Smoothing of All Shapes and Sizescited by 146
  11. 2020Perceptual Adversarial Robustness: Defense Against Unseen Threat Modelscited by 137
  12. 2018Parametric Noise Injection: Trainable Randomness to Improve Deep Neural Network Robustness against Adversarial Attackcited by 123
  13. 2019Adversarial Examples Are a Natural Consequence of Test Error in Noisecited by 114
  14. 2018Decoupling Direction and Norm for Efficient Gradient-Based L2 Adversarial Attacks and Defensescited by 109
  15. 2019Adversarial Robustness Against the Union of Multiple Perturbation Modelscited by 88
  16. 2019When NAS Meets Robustness: In Search of Robust Architectures against Adversarial Attackscited by 87
  17. 2019Unrestricted Adversarial Examples via Semantic Manipulationcited by 84
  18. 2019Functional Adversarial Attackscited by 82
  19. 2019Model Compression with Adversarial Robustness: A Unified Optimization Frameworkcited by 71
  20. 2019Efficient Adversarial Training with Transferable Adversarial Examplescited by 54
  21. 2020Triple Wins: Boosting Accuracy, Robustness and Efficiency Together by Enabling Input-Adaptive Inferencecited by 45
  22. 2020Single-step Adversarial training with Dropout Schedulingcited by 40
  23. 2020Randomization matters. How to defend against strong adversarial attackscited by 37
  24. 2019Theoretical evidence for adversarial robustness through randomizationcited by 33
  25. 2019Intriguing properties of adversarial training at scalecited by 33
  26. 2019Transfer of Adversarial Robustness Between Perturbation Typescited by 25
  27. 2019Quantifying Perceptual Distortion of Adversarial Examplescited by 10

Built on by

Not cited yet.