Towards a General-Purpose Homomorphically Encrypted Microprocessor
Department of Electrical Engineering Seminar Series
Michail (Mihalis) Maniatakos
Assistant Professor of Electrical and Computer Engineering
New York University (NYU) Abu Dhabi
"Towards a General-Purpose Homomorphically Encrypted Microprocessor"
Abstract: Modern microprocessors have always been designed for maximizing performance. A plethora of recent vulnerability disclosures, however, serve as a painful reminder that all these years of performance gains through microarchitectural improvements were not exactly "free lunch". Evidently, end-points cannot be trusted; as long as data gets decrypted to be processed in the microprocessor pipeline, it can leak. While cryptography has been successfully used to solve data-in-transit (e.g., HTTPS) and data-at-rest (e.g., AES encrypted hard disks) concerns, data-in-use protection in regular CPUs remains unsolved. Fully homomorphic encryption has emerged as a prominent solution, but its performance and applicability render it unsuitable for an across the board solution. In this talk, we will present the major challenges towards developing a homomorphically encrypted processor, starting from the development of a minimal abstract machine that can compute on encrypted data, to an ASIC implementation of partially homomorphic encrypted microprocessor and the introduction of a framework for compiling C++ programs to their homomorphically encrypted counterparts
Bio: Michail (Mihalis) Maniatakos is an Assistant Professor of Electrical and Computer Engineering at New York University (NYU) Abu Dhabi. He is the Director of the MoMA Laboratory (nyuad.nyu.edu/momalab). He received his Ph.D. in Electrical Engineering from Yale University. His research interests include encrypted data computation and industrial control systems security, and has been funded by various industry partners, DARPA, and ONR.
Hosted by: Professor Jakub Szefer
Tuesday, June 4, 2019
Mann Student Center- 4:00 PM