A game-based approximate verification of deep neural networks with provable guarantees

• Published in: Theoretical computer science Vol. 807; pp. 298 - 329
• Main Authors: Wu, Min, Wicker, Matthew, Ruan, Wenjie, Huang, Xiaowei, Kwiatkowska, Marta
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 06.02.2020
• Subjects: Adversarial examples; Automated verification; Deep neural networks; Two-player game; Two-player game; Adversarial examples; Deep neural networks; Automated verification

Despite the improved accuracy of deep neural networks, the discovery of adversarial examples has raised serious safety concerns. In this paper, we study two variants of pointwise robustness, the maximum safe radius problem, which for a given input sample computes the minimum distance to an adversari...