Verification for Robust neural networks
through Abstraction, Partitioning, and Attack methods

VeRAPAk is a highly versatile, refutation-based framework designed for the adversarial robustness verification of deep neural classification networks.