Robust Detection of Out-of-Distribution and Adversarial Inputs for Deep Neural Networks
Robust methods to accurately identify both out-of-distribution and adversarially perturbed inputs, even when they are designed to evade the out-of-distribution detector.