Timezone: »

 
Towards Out-of-Distribution Adversarial Robustness
Adam Ibrahim · Charles Guille-Escuret · Ioannis Mitliagkas · Irina Rish · David Krueger · Pouya Bashivan

@
in 2nd ICML Workshop on New Frontiers in Adversarial Machine Learning »
Event URL: https://openreview.net/forum?id=JDlq499bat »
Adversarial robustness continues to be a major challenge for deep learning. A core issue is that robustness to one type of attack often fails to transfer to other attacks. While prior work establishes a theoretical trade-off in robustness against different $L_p$ norms, we show that there is potential for improvement against many commonly used attacks by adopting a domain generalisation approach.Concretely, we treat each type of attack as a domain, and apply the Risk Extrapolation method (REx), which promotes similar levels of robustness against all training attacks. Compared to existing methods, we obtain similar or superior worst-case adversarial robustness on attacks seen during training. Moreover, we achieve superior performance on families or tunings of attacks only encountered at test time. On ensembles of attacks, our approach improves the accuracy from 3.4\% with the best existing baseline to 25.9\% on MNIST, and from 16.9\% to 23.5\% on CIFAR10.
Keywords: domain generalization · robustness · adversarial · OOD · REx

Author Information

Adam Ibrahim (Mila, University of Montreal)
Charles Guille-Escuret (Mila, Université de Montréal)
Ioannis Mitliagkas (MILA, UdeM)
Irina Rish (MILA / Université de Montréal h)
David Krueger (MILA (University of Montreal))
Pouya Bashivan (McGill University)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors