Adversarial Robustness via Label-Smoothing

Abstract

We study Label-Smoothing as a means for improving adversarial robustness of supervised deep-learning models. After establishing a thorough and unified framework, we propose several variations to this general method: adversarial, Boltzmann and second-best Label-Smoothing methods, and we explain how to construct your own one. On various datasets (MNIST, CIFAR10, SVHN) and models (linear models, MLPs, LeNet, ResNet), we show that Label-Smoothing in general improves adversarial robustness against a variety of attacks (FGSM, BIM, DeepFool, Carlini-Wagner) by better taking account of the dataset geometry. The proposed Label-Smoothing methods have two main advantages: they can be implemented as a modified cross-entropy loss, thus do not require any modifications of the network architecture nor do they lead to increased training times, and they improve both standard and adversarial accuracy.

References in corpus (8)

• The Space of Transferable Adversarial Examples
• Robustness May Be at Odds with Accuracy
• Adversarial Examples: Opportunities and Challenges
• DARTS: Deceiving Autonomous Cars with Toxic Signs
• A Boundary Tilting Persepective on the Phenomenon of Adversarial Examples
• Robustness of classifiers: from adversarial to random noise
• On the Effectiveness of Defensive Distillation
• Label Smoothing and Logit Squeezing: A Replacement for Adversarial Training?

Cited by in corpus (4)

• Towards Understanding Label Smoothing
• Defending Against Adversarial Attacks by Suppressing the Largest Eigenvalue of Fisher Information Matrix
• Lower Bounds on Cross-Entropy Loss in the Presence of Test-time Adversaries
• Beyond Categorical Label Representations for Image Classification