On Secure Distributed Linearly Separable Computation

Distributed linearly separable computation, where a user asks some distributed servers to compute function, was recently formulated by the same authors and aims alleviate bottlenecks of stragglers communication cost in computation. The data center assigns subset input datasets each server an uncoded manner, computes coded packets on assigned datasets, which are then sent user. should recover task function from answers servers, such that effect could be tolerated. In this paper, we formulate novel secure framework for aim let only retrieve desired without obtaining any other information about even if it receives all servers. order preserve security common randomness variable independent introduced into transmission. We show non-secure linear-coding based computing scheme original computation problem, can made increasing (number symbols receive). Then focus case server) is minimum minimize size (i.e., size) system while achieving optimal cost. first propose theoretic converse bound size. schemes two well-known assignments, namely fractional repetition assignment xmlns:xlink="http://www.w3.org/1999/xlink">cyclic . These subject using these assignments. Motivated observation general limitation size, with assignment, strictly outperforms above schemes. Some additional optimality results also obtained.