A Recipe for Improved Certifiable Robustness

Recent studies have highlighted the potential of Lipschitz-based methods for training certifiably robust neural networks against adversarial attacks. A key challenge, supported both theoretically and empirically, is that robustness demands greater network capacity and more data than standard trainin...

Ausführliche Beschreibung

Bibliographische Detailangaben

Veröffentlicht in:	arXiv.org 2024-06
Hauptverfasser:	Hu, Kai, Leino, Klas, Wang, Zifan, Fredrikson, Matt
Format:	Artikel
Sprache:	eng
Schlagworte:	Certification Data augmentation Design optimization Perturbation Robustness State of the art
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!