A Game Theoretic Analysis of Additive Adversarial Attacks and Defenses

Abstract

Research in adversarial learning follows a cat and mouse game between attackers and defenders where attacks are proposed, they are mitigated by new defenses, and subsequently new attacks are proposed that break earlier defenses, and so on. However, it has remained unclear as to whether there are conditions under which no better attacks or defenses can be proposed. In this paper, we propose a game-theoretic framework for studying attacks and defenses which exist in equilibrium. Under a locally linear decision boundary model for the underlying binary classifier, we prove that the Fast Gradient Method attack and the Randomized Smoothing defense form a Nash Equilibrium. We then show how this equilibrium defense can be approximated given finitely many samples from a data-generating distribution, and derive a generalization bound for the performance of our approximation.

arXiv commentAccepted at Neural Information Processing Systems (NeurIPS) 2020

References in corpus (18)

• Feature Squeezing: Detecting Adversarial Examples in Deep Neural Networks
• ZOO: Zeroth Order Optimization based Black-box Attacks to Deep Neural Networks without Training Substitute Models
• Transferability in Machine Learning: from Phenomena to Black-Box Attacks using Adversarial Samples
• Certified Adversarial Robustness via Randomized Smoothing
• Distributional Smoothing with Virtual Adversarial Training
• PixelDefend: Leveraging Generative Models to Understand and Defend against Adversarial Examples
• Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulation
• A study of the effect of JPG compression on adversarial images
• Exploring the Landscape of Spatial Robustness
• Adversarial Robustness through Local Linearization
• Certified Adversarial Robustness with Additive Noise
• Lipschitz-Margin Training: Scalable Certification of Perturbation Invariance for Deep Neural Networks
• Provably Minimally-Distorted Adversarial Examples
• Output Range Analysis for Deep Neural Networks
• Tight Certificates of Adversarial Robustness for Randomly Smoothed Classifiers
• Robustness of classifiers to universal perturbations: a geometric perspective
• Towards Robust, Locally Linear Deep Networks
• Nonzero-sum Adversarial Hypothesis Testing Games