Improving Adversarial Robustness via Probabilistically Compact Loss with Logit Constraints
This work aims to improve the adversarial robustness of CNNs, which is a critical problem for the deployment of AI systems in security-sensitive applications, by offering an incremental improvement to existing loss functions.
The paper addresses the vulnerability of Convolutional Neural Networks (CNNs) to adversarial samples, which often leads to misclassification into the most probable false classes. The authors propose a Probabilistically Compact (PC) loss with logit constraints, designed to enlarge probability gaps between true and false classes and prevent these gaps from being reduced by small perturbations.
Convolutional neural networks (CNNs) have achieved state-of-the-art performance on various tasks in computer vision. However, recent studies demonstrate that these models are vulnerable to carefully crafted adversarial samples and suffer from a significant performance drop when predicting them. Many methods have been proposed to improve adversarial robustness (e.g., adversarial training and new loss functions to learn adversarially robust feature representations). Here we offer a unique insight into the predictive behavior of CNNs that they tend to misclassify adversarial samples into the most probable false classes. This inspires us to propose a new Probabilistically Compact (PC) loss with logit constraints which can be used as a drop-in replacement for cross-entropy (CE) loss to improve CNN's adversarial robustness. Specifically, PC loss enlarges the probability gaps between true class and false classes meanwhile the logit constraints prevent the gaps from being melted by a small perturbation. We extensively compare our method with the state-of-the-art using large scale datasets under both white-box and black-box attacks to demonstrate its effectiveness. The source codes are available from the following url: https://github.com/xinli0928/PC-LC.