UMass AI&Sec SP'25 Seminar: Ryan McKenna (Google), Private Analytics and Learning at Google
Content
Speaker
Abstract
In this talk, I will give a broad overview of how we think about the many dimensions of data privacy at Google, discuss some of the private analytics and learning problems we’ve faced here, and go over some of the research we have done in this space to solve these problems, as well as promising open research directions. To that end, I will talk about how we used our federated analytics platform to derive environmental insights from on-device location data and how Gboard trains their next-word prediction model with strong differential privacy guarantees and without centralized data collection. I will then overview a body of research we have done on improving DP-SGD, the most widely used mechanism for training machine learning models with differential privacy.
Bio
Ryan McKenna is a Research Scientist at Google, specializing in differential privacy, graphical models, and machine learning. He earned his Ph.D. in Computer Science from the University of Massachusetts Amherst in 2022, where he was advised by Professors Gerome Miklau and Daniel Sheldon. His research focuses on developing privacy-preserving data analytics techniques, and he has published in leading conferences such as ICML and NeurIPS. Notably, Ryan was part of the winning team in the 2018 NIST Differential Privacy Synthetic Data Challenge.
Host