College of Information and Computer Sciences (CICS) doctoral student Ryan McKenna recently won first place and a $25,000 prize in the National Institute of Standards and Technology's (NIST) 2019 Differential Privacy Synthetic Data Challenge.
The goal of the challenge was to accelerate the development of high-quality, differentially private data release tools. Competitors participated in three marathon coding matches with the goal of designing and implementing a synthetic data generation algorithm, and in addition prove that it satisfied differential privacy.
McKenna’s winning algorithm used the Gaussian mechanism to obtain noisy answers to a carefully selected set of counting queries, then used graphical models to find the data distribution that best explained those noisy observations and sample synthetic data.
McKenna currently works in the CICS DREAM Lab (Data systems Research for Exploration, Analytics, and Modeling) with Professor Gerome Miklau, where his research focuses on data privacy and machine learning. Last year, he took third place in the NIST Differential Privacy Synthetic Data Challenge in its inaugural competition.
His complete winning algorithm can be found on Github.