1 paper · 1 filter
Philip Sosnin, Jodie Knapp, Fraser Kennedy +2
This work introduces a verification framework that provides both sound and complete guarantees for data poisoning attacks during neural network training. We formulate adversarial d…