It took the right pieces of data coming together in one place to be able to accurately tell which factors about a vulnerability are descriptive of the likelihood exploitation. Next up: predictions. We use Amazon Web Services new Machine Learning platform to train a binary classifier for vulnerabilities, allowing us to predict whether or not a vulnerability will become exploitable. We enrich the data with more specific, non - definitional level data as pioneered by Allodi, Massacci, Geer, and the present speaker. 500 million live vulnerabilities and their associated close rates inform the epidemiological data, as well as "in the wild" threat data from Alienvault's OTX, Secureworks' CTU, Reversing Labs, and ISC SANS. We use 70% of the national vulnerability database as our training dataset, and generate over 20,000 predictions on the remainder of the vulnerabilities. We measure specificity and sensitivity, positive predictive value, false positive and false negative rates, and arrive at an optimal decision cutoff for the problem.

Michael Roytman is the Chief Data Scientist at Kenna Security, and has been selected to speak at some of the top security conferences in the world, including RSA, SOURCE, Bsides, Metricon and SIRAcon. His work focuses on cybersecurity data science and Bayesian algorithms. He is also a technical advisor in the humanitarian space, having worked with Doctors Without Borders, The World Health Organization, and the UN. He is the cofounder and a board member of Dharma Platform, for which he landed on the 2017 Forbes 30 under 30 list. He holds an M.S. in Operations Research from Georgia Tech, and his home in Chicago houses an industrial-scale coffee roasting operation.

Back to Converge 2017 video list