Cross-Input Certified Training for Universal Perturbations
Changming Xu, Gagandeep Singh
TL;DR
This work proposes a novel method, CITRUS, for certified training of networks robust against UAP attackers, and shows in an extensive evaluation that this method outperforms traditional certified training methods on standard accuracy and achieves SOTA performance on the more practical certified UAP accuracy metric.
Abstract
Existing work in trustworthy machine learning primarily focuses on single-input adversarial perturbations. In many real-world attack scenarios, input-agnostic adversarial attacks, e.g. universal adversarial perturbations (UAPs), are much more feasible. Current certified training methods train models robust to single-input perturbations but achieve suboptimal clean and UAP accuracy, thereby limiting their applicability in practical applications. We propose a novel method, CITRUS, for certified training of networks robust against UAP attackers. We show in an extensive evaluation across different datasets, architectures, and perturbation magnitudes that our method outperforms traditional certified training methods on standard accuracy (up to 10.3\%) and achieves SOTA performance on the more practical certified UAP accuracy metric.