Information-Theoretically Private Matrix Multiplication From MDS-Coded Storage

We study two problems of private matrix multiplication, over a distributed computing system consisting master node, and multiple servers that collectively store family public matrices using Maximum-Distance-Separable (MDS) codes. In the first problem Private Secure Matrix Multiplication (PSMM) from colluding servers, intends to compute product its confidential A with target stored on without revealing any information about index some servers. second Fully (FPMM) is also selected another at in MDS form. this case, indices should both be kept develop novel strategies for PSMM FPMM problems, which simultaneously guarantee information-theoretic data/index privacy computation correctness. compare proposed strategy previous weaker (non-colluding servers), demonstrate substantial improvements terms communication overheads. Moreover, compared baseline uses idea Information Retrieval (PIR) directly retrieve desired significantly reduces storage overhead, but slightly incurs large