This deliverable presents the results from two studies on adversarial machine learning (ML). Their goal is to examine the nature of this risk for cybersecurity-related applications of ML and to highlight the importance of training models in adversarial settings.