17
1

RAMP: Boosting Adversarial Robustness Against Multiple lpl_p Perturbations

Abstract

There is considerable work on improving robustness against adversarial attacks bounded by a single lpl_p norm using adversarial training (AT). However, the multiple-norm robustness (union accuracy) of AT models is still low. We observe that simultaneously obtaining good union and clean accuracy is hard since there are tradeoffs between robustness against multiple lpl_p perturbations, and accuracy/robustness/efficiency. By analyzing the tradeoffs from the lens of distribution shifts, we identify the key tradeoff pair among lpl_p attacks to boost efficiency and design a logit pairing loss to improve the union accuracy. Next, we connect natural training with AT via gradient projection, to find and incorporate useful information from natural training into AT, which moderates the accuracy/robustness tradeoff. Combining our contributions, we propose a framework called \textbf{RAMP}, to boost the robustness against multiple lpl_p perturbations. We show \textbf{RAMP} can be easily adapted for both robust fine-tuning and full AT. For robust fine-tuning, \textbf{RAMP} obtains a union accuracy up to 53.5%53.5\% on CIFAR-10, and 29.7%29.7\% on ImageNet. For training from scratch, \textbf{RAMP} achieves SOTA union accuracy of 44.6%44.6\% and relatively good clean accuracy of 81.2%81.2\% on ResNet-18 against AutoAttack on CIFAR-10.

View on arXiv
Comments on this paper