Jacobian Regularization for Mitigating Universal Adversarial Perturbations
Authors:
- Kenneth T. CoKCKenneth T. Co
- David Martinez RegoDRDavid Martinez Rego
- Emil C. LupuELEmil C. Lupu
Preprint
Abstract
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.Keywords
All Related Versions
- Version 1, 2021-04-21, ArXiv
- Version 2, 2021-09-13, ArXiv
- Published version: , 202.
This publication has 11 references indexed in Scilit:
- Shadow-Catcher: Looking into Shadows to Detect Ghost Objects in Autonomous Vehicle 3D Sensing2021
- Universal Adversarial Robustness of Texture and Shape-Biased Models2021
- Object Removal Attacks on LiDAR-based 3D Object Detectors2021
- Physically Realizable Adversarial Examples for LiDAR Object Detection2020
- Universal Adversarial TrainingProceedings of the AAAI Conference on Artificial Intelligence, 2020
- Procedural Noise Adversarial Examples for Black-Box Attacks on Deep Convolutional Networks2019
- AdVersarial2019
- Robust Large Margin Deep Neural NetworksIEEE Transactions on Signal Processing, 2017
- Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research GroupsIEEE Signal Processing Magazine, 2012
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998