Download as iCal file


How to Secretly Multiply Matrices: from Secret Sharing to Coded Computation

Salim El Rouayheb (Rutgers University ECE Department)

Location:  Hill 705 and Zoom
Date & time: Friday, 01 December 2023 at 2:45PM - 3:30PM

Abstract: I will talk about  the problem of Secure Distributed Matrix Multiplication (SDMM) in which a data owner wishes to compute the product of two matrices using the assistance of honest but curious, and possibly non-responsive, helpers. Building on Shamir’s classical results on secret sharing, I will establish a connection between SDMM  codes with low communication cost to a new combinatorial object that we call Additive Degree Table (ADT).  By studying ADTs, we devise a new parameterized family of codes for SDMM that we call GASP (Gap Additive Secure Polynomial) Codes. We also derive lower bounds and prove that GASP codes are optimal in certain regimes. This is joint work with Rafael D'Oliveira, Daniel Heinlein, and David Karpuk.