Characterizing the Optimal 0-1 Loss for Multi-class Classification with a Test-time Attacker. (arXiv:2302.10722v1 [cs.LG])

Finding classifiers robust to adversarial examples is critical for their safe
deployment. Determining the robustness of the best possible classifier under a
given threat model for a given data distribution and comparing it to that
achieved by state-of-the-art training methods is thus an important diagnostic
tool. In this paper, we find achievable information-theoretic lower bounds on
loss in the presence of a test-time attacker for multi-class classifiers on any
discrete dataset. We provide a general framework for finding the optimal …

adversarial art class classification critical data deployment distribution find important information loss robustness safe state test threat threat model tool training under