ACE Seminar: Privacy-Preserving Data Analysis: Proofs, Algorithms, and Systems

Speaker: Adria Gascon

Date/Time: 25-Apr-2019, 16:00 UTC

Venue: Roberts 309



A fully-fledged privacy-aware data analysis pipeline must take into account different notions of privacy and their interplays. In this talk we will introduce new protocols for secure Machine Learning (ML) in the Multi-Party Computation (MPC) setting, and discuss their interplays with Differential Privacy (DP). More concretely, I will present variants of neural networks with training procedures that are amenable to secure computation. Moreover, we will see how, although generally speaking MPC and DP address complementary aspects of computing privately, a careful combination leads to hybrid definitions with satisfactory practical consequences in terms of both running time and accuracy.


Adria Gascon is a computer scientist with research interests in formal languages and compression, automated reasoning, cryptography, and machine learning. He earned his PhD from the Technical University of Catalonia, and has held positions at SRI International and the University of Edinburgh.

Add to Calendar

This page was last modified on 27 Mar 2014.