Tight Second-Order Certificates for Randomized Smoothing

Abstract

Randomized smoothing is a popular way of providing robustness guarantees against adversarial attacks: randomly-smoothed functions have a universal Lipschitz-like bound, allowing for robustness certificates to be easily computed. In this work, we show that there also exists a universal curvature-like bound for Gaussian random smoothing: given the exact value and gradient of a smoothed function, we compute a lower bound on the distance of a point to its closest adversarial example, called the Second-order Smoothing (SoS) robustness certificate. In addition to proving the correctness of this novel certificate, we show that SoS certificates are realizable and therefore tight. Interestingly, we show that the maximum achievable benefits, in terms of certified robustness, from using the additional information of the gradient norm are relatively small: because our bounds are tight, this is a fundamental negative result. The gain of SoS certificates further diminishes if we consider the estimation error of the gradient norms, for which we have developed an estimator. We therefore additionally develop a variant of Gaussian smoothing, called Gaussian dipole smoothing, which provides similar bounds to randomized smoothing with gradient information, but with much-improved sample efficiency. This allows us to achieve (marginally) improved robustness certificates on high-dimensional datasets such as CIFAR-10 and ImageNet. Code is available at https://github.com/alevine0/smoothing_second_order.

Updated to reference and reflect results of concurrent work, Mohapatra et al., which shows that second-order certificates are possible using sqrt(d) samples

References in corpus (10)

• Provable defenses against adversarial examples via the convex outer adversarial polytope
• Certified Adversarial Robustness via Randomized Smoothing
• Evaluating Robustness of Neural Networks with Mixed Integer Programming
• On the Effectiveness of Interval Bound Propagation for Training Verifiably Robust Models
• Provably Robust Deep Learning via Adversarially Trained Smoothed Classifiers
• Semidefinite relaxations for certifying robustness to adversarial examples
• Certified Adversarial Robustness with Additive Noise
• Provably Minimally-Distorted Adversarial Examples
• Preventing Gradient Attenuation in Lipschitz Constrained Convolutional Networks
• Certifiably Robust Interpretation in Deep Learning