Rethinking Uncertainty in Deep Learning: Whether and How it Improves Robustness
This work provides a principled way to enhance the robustness of deep learning models for practitioners and researchers working on adversarial defense, addressing the limitations of adversarial training.
This paper investigates the role of uncertainty promotion regularizers like entropy maximization (EntM) and label smoothing (LS) in improving the robustness of deep neural networks against adversarial attacks. It finds that while these regularizers alone offer limited protection against strong attacks, they significantly complement adversarial training (AT), leading to consistent performance improvements on clean examples and under various attacks, particularly those with large perturbations.
Deep neural networks (DNNs) are known to be prone to adversarial attacks, for which many remedies are proposed. While adversarial training (AT) is regarded as the most robust defense, it suffers from poor performance both on clean examples and under other types of attacks, e.g. attacks with larger perturbations. Meanwhile, regularizers that encourage uncertain outputs, such as entropy maximization (EntM) and label smoothing (LS) can maintain accuracy on clean examples and improve performance under weak attacks, yet their ability to defend against strong attacks is still in doubt. In this paper, we revisit uncertainty promotion regularizers, including EntM and LS, in the field of adversarial learning. We show that EntM and LS alone provide robustness only under small perturbations. Contrarily, we show that uncertainty promotion regularizers complement AT in a principled manner, consistently improving performance on both clean examples and under various attacks, especially attacks with large perturbations. We further analyze how uncertainty promotion regularizers enhance the performance of AT from the perspective of Jacobian matrices $\nabla_X f(X;θ)$, and find out that EntM effectively shrinks the norm of Jacobian matrices and hence promotes robustness.