Fully Private Coded Matrix Multiplication From Colluding Workers

Minchul Kim, Heecheol Yang, Jungwoo Lee

Abstract: Recently, coded computation has been used to reduce the completion time in distributed computing by mitigating straggler effects with erasure codes. As a variation of coded computation, fully private coded matrix multiplication (FPCMM) has been proposed to preserve a master's privacy in a scenario where the master wants to obtain a matrix multiplication result from the libraries which are shared at the workers, while concealing both of the two indices of the desired matrices from each worker. In this letter, we propose a new FPCMM scheme to keep a master's privacy from up to T colluding workers. Furthermore, we compare the performance of our scheme with those of the existing private coded matrix multiplication schemes for non-colluding workers and for concealing the index of a single desired matrix.