Representation of white- and black-box adversarial examples in deep neural networks and humans: A functional magnetic resonance imaging study.