Limitations of the Lipschitz constant as a defense against adversarial examples

Abstract

Several recent papers have discussed utilizing Lipschitz constants to limit the susceptibility of neural networks to adversarial examples. We analyze recently proposed methods for computing the Lipschitz constant. We show that the Lipschitz constant may indeed enable adversarially robust neural networks. However, the methods currently employed for computing it suffer from theoretical and practical limitations. We argue that addressing this shortcoming is a promising direction for future research into certified adversarial defenses.

References in corpus (1)

• Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples

Cited by in corpus (8)

• Sorting out Lipschitz function approximation
• Defending Against Adversarial Examples with K-Nearest Neighbor
• Adversarial Examples on Object Recognition: A Comprehensive Survey
• Universal Lipschitz Approximation in Bounded Depth Neural Networks
• Some open questions on morphological operators and representations in the deep learning era
• Lipschitz Networks and Distributional Robustness
• Towards Certifying L-infinity Robustness using Neural Networks with L-inf-dist Neurons
• Global Adversarial Attacks for Assessing Deep Learning Robustness