Statistical Classification and Computer Security

During the last couple of years we have been addressing classification problems in computer security applications. In particular, we have focused on three key items: (1) evaluation metrics for classifiers in adversarial environments, (2) the design of optimal classifiers against adaptive adversaries, and (3) voting algorithms for the combination of multiple classifiers. Our results have been published in computer security [1, 2, 3, 4] and machine learning [5, 6, 7, 8] venues. In this abstract we summarize our results, discuss our current work, and mention some other open problems in the intersection of machine learning and computer security.