Secure Batch Matrix Multiplication From Grouping Lagrange Encoding
Jinbao Zhu, Xiaohu Tang · IEEE Communications Letters · 2020
In this letter, the problem of distributed Secure Batch Matrix Multiplication (SBMM) is studied, where a user wishes to compute the pairwise products of two batches of massive matrices$\mathbf {A}$and$\mathbf {B}$generated by two external source nodes, with the aid of$N$distributed servers. The security for data matrices$\mathbf {A}$(resp.$\mathbf {B}$) is guaranteed against any group of up to$X_{\mathbf {A}}$(resp.$X_{\mathbf {B}}$) colluding servers. As a result, a computation strategy is presented to characterize the trade-off between recovery threshold, system cost and system complexity, based on grouping Lagrange encoding, which unifies and improves the previous strategies for SBMM.