Robustness Bounds on the Successful Adversarial Examples: Theory and Practice

arXiv:2403.01896v1 Announce Type: new
Abstract: Adversarial example (AE) is an attack method for machine learning, which is crafted by adding imperceptible perturbation to the data inducing misclassification. In the current paper, we investigated the upper bound of the probability of successful AEs based on the Gaussian Process (GP) classification. We proved a new upper bound that depends on AE's perturbation norm, the kernel function used in GP, and the distance of the closest pair with different labels in the training …

abstract adversarial adversarial examples arxiv classification cs.cr cs.lg current data example examples machine machine learning paper practice probability process robustness stat.ml theory type