Uncovering the Limits of Adversarial Training against Norm-Bounded Adversarial Examples
This work addresses the critical issue of enhancing model robustness against adversarial attacks for AI safety, representing an incremental advance through systematic optimization of existing methods.
The paper tackles the problem of improving adversarial robustness in deep neural networks by exploring the limits of adversarial training, achieving state-of-the-art results with large improvements in accuracy under attack on CIFAR-10 and CIFAR-100 datasets, such as 65.88% accuracy against l∞ perturbations on CIFAR-10 with unlabeled data.
Adversarial training and its variants have become de facto standards for learning robust deep neural networks. In this paper, we explore the landscape around adversarial training in a bid to uncover its limits. We systematically study the effect of different training losses, model sizes, activation functions, the addition of unlabeled data (through pseudo-labeling) and other factors on adversarial robustness. We discover that it is possible to train robust models that go well beyond state-of-the-art results by combining larger models, Swish/SiLU activations and model weight averaging. We demonstrate large improvements on CIFAR-10 and CIFAR-100 against $\ell_\infty$ and $\ell_2$ norm-bounded perturbations of size $8/255$ and $128/255$, respectively. In the setting with additional unlabeled data, we obtain an accuracy under attack of 65.88% against $\ell_\infty$ perturbations of size $8/255$ on CIFAR-10 (+6.35% with respect to prior art). Without additional data, we obtain an accuracy under attack of 57.20% (+3.46%). To test the generality of our findings and without any additional modifications, we obtain an accuracy under attack of 80.53% (+7.62%) against $\ell_2$ perturbations of size $128/255$ on CIFAR-10, and of 36.88% (+8.46%) against $\ell_\infty$ perturbations of size $8/255$ on CIFAR-100. All models are available at https://github.com/deepmind/deepmind-research/tree/master/adversarial_robustness.