Mariana Raykova
April 24, 2024 at 11:00 AM on Zoom / Soda Hall
Advances (And Challenges) in Secure Aggregation
Abstract: Systems for private analytics and federated learning rely on aggregation of distributed data generated across many user devices. In this talk we will discuss cryptographic techniques for distributed secure aggregation in a few different models and the applications that have adopted them. We will highlight some new technical results in this area including how we can use lattices instead of PRGs to improve efficiency and how we can achieve input norm bounding guarantees in the setting of single server aggregation. We will discuss the challenges of differentially private aggregation of sparse inputs in both single and multi-server settings and we will present a new construction in the two server model.
Bio: Mariana is a Research Scientist in the Private Computing Group at Google. Her research includes work in the areas of secure computation, oblivious data structures, zero knowledge, and verifiable computation, obfuscation. She received her PhD from the Computer Science Department of Columbia University and was co-advised by Tal Malkin and Steve Bellovin. After her PhD, she spent a year as a postdoc at the Cryptography Group at IBM Research Watson. She was a Research Scientist at the Computer Science Laboratory at SRI International between 2013 and 2015. Following that, she was an Assistant Professor at the Department of Computer Science at Yale University between 2016 and 2018.