mlsanityData to 14 Oct 2026

Jacobian Regularization for Mitigating Universal Adversarial Perturbations

arXiv 2104.10459 ↗First posted 21 Apr 2021cs.LG cs.AI cs.CRcited by 4

Not on the map. Why?

Abstract (arXiv)

Universal Adversarial Perturbations (UAPs) are input perturbations that can fool a neural network on large sets of data. They are a class of attacks that represents a significant threat as they facilitate realistic, practical, and low-cost attacks on neural networks. In this work, we derive upper bounds for the effectiveness of UAPs based on norms of data-dependent Jacobians. We empirically verify that Jacobian regularization greatly increases model robustness to UAPs by up to four times whilst maintaining clean performance. Our theoretical analysis also allows us to formulate a metric for the strength of shared adversarial perturbations between pairs of inputs. We apply this metric to benchmark datasets and show that it is highly correlated with the actual observed robustness. This suggests that realistic and practical universal attacks can be reliably mitigated without sacrificing clean accuracy, which shows promise for the robustness of machine learning systems.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Builds on 25 of its 31 references that are in the map. First posted on arXiv 21 Apr 2021.

Builds on

25 of 31 references.

References · 25

  1. 2013Intriguing properties of neural networkscited by 5,155
  2. 2015You Only Look Once: Unified, Real-Time Object Detectioncited by 4,852
  3. 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,349
  4. 2016Universal adversarial perturbationscited by 784
  5. 2017Evasion Attacks against Machine Learning at Test Timecited by 689
  6. 2017Adversarial Patchcited by 517
  7. 2019Fooling automated surveillance cameras: adversarial patches to attack person detectioncited by 268
  8. 2018Sensitivity and Generalization in Neural Networks: an Empirical Studycited by 189
  9. 2018Physical Adversarial Examples for Object Detectorscited by 163
  10. 2019Adversarial Sensor Attack on LiDAR-based Perception in Autonomous Drivingcited by 152
  11. 2020Physically Realizable Adversarial Examples for LiDAR Object Detectioncited by 137
  12. 2019Robust Learning with Jacobian Regularizationcited by 129
13 more
  1. 2016Robust Large Margin Deep Neural Networkscited by 118
  2. 2018Improving DNN Robustness to Adversarial Attacks using Jacobian Regularizationcited by 88
  3. 2018Universal Adversarial Trainingcited by 87
  4. 2019Adversarial Objects Against LiDAR-Based Autonomous Driving Systemscited by 63
  5. 2017Art of singular vectors and universal adversarial perturbationscited by 49
  6. 2017Gradient Regularization Improves Accuracy of Discriminative Modelscited by 26
  7. 2018Defending Against Universal Perturbations With Shared Adversarial Trainingcited by 25
  8. 2018Procedural Noise Adversarial Examples for Black-Box Attacks on Deep Convolutional Networkscited by 21
  9. 2021Object Removal Attacks on LiDAR-based 3D Object Detectorscited by 16
  10. 2018AdVersarial: Perceptual Ad Blocking meets Adversarial Machine Learningcited by 14
  11. 2019Universal Adversarial Robustness of Texture and Shape-Biased Modelscited by 8
  12. 2019Sensitivity of Deep Convolutional Networks to Gabor Noisecited by 3
  13. 2020Shadow-Catcher: Looking Into Shadows to Detect Ghost Objects in Autonomous Vehicle 3D Sensingcited by 2

Built on by

Cited by year

Citing papers per year: 2021: 1; 2022: 2; 2023: 1.

4 papers.

  1. 2021Real-time Detection of Practical Universal Adversarial Perturbationscited by 6
  2. 2022Towards an Awareness of Time Series Anomaly Detection Models' Adversarial Vulnerabilitycited by 5
  3. 2022Jacobian Ensembles Improve Robustness Trade-offs to Adversarial Attackscited by 4
  4. 2023Bridging Optimal Transport and Jacobian Regularization by Optimal Trajectory for Enhanced Adversarial Defensecited by 0