Security and Privacy in Cache-Aided Linear Function Retrieval for Multi-access Coded Caching

TL;DR

This paper extends multi-access coded caching to include linear function retrieval, security, and privacy, proposing schemes that are near-optimal under certain memory conditions in a combinatorial topology network.

Contribution

It introduces secure and private linear function retrieval schemes for multi-access caching, generalizing previous models and achieving near-optimal rates under specific conditions.

Findings

Proposed secure, private, and improved schemes for linear function retrieval.

Schemes recover previous results as special cases.

Achieves near-optimal rates within a constant factor under certain memory conditions.

Abstract

A multi-access network consisting of $N$ files, $C$ caches, $K$ users with each user having access to a unique set of $r$ caches has been introduced recently by Muralidhar et al. ("Maddah-Ali-Niesen Scheme for Multi-access Coded Caching," in \textit{Proc. ITW}, 2021). It considers Single File Retrieval (SFR) i.e, each user demands an arbitrary file from the server. It proposes a coded caching scheme which was shown to be optimal under the assumption of uncoded placement by Brunero and Elia ("Fundamental Limits of Combinatorial Multi-Access Caching" in {\textit{arXiv:2110.07426} }). The above multi-access network is referred to as combinatorial topology which is considered in this work with three additional features : a) Linear Function Retrieval (LFR) i.e., each user is interested in retrieving an arbitrary linear combination of files in the server's library; b) Security i.e., the content of the library must be kept secure from an eavesdropper who obtains the signal sent by the server; c) Privacy i.e., each user can only get its required file and can not get any information about the demands of other users. Achievable Secure, Private LFR (SP-LFR) scheme, Secure LFR (S-LFR) scheme and Improved S-LFR scheme are proposed. As special cases, our work recovers some of the results by Yan and Tuninetti ("Key Superposition Simultaneously Achieves Security and Privacy in Cache-Aided Linear Function Retrieval," in \textit{Trans. Inf. Forensics and Security}, 2021") and Sengupta et al.("Fundamental limits of caching with secure delivery," in \textit{Trans. Inf. Forensics and Security}, 2015). At a memory point, $M=\frac{r\binom{C}{r}}{C}$, the SP-LFR scheme is within a constant multiplicative factor from the optimal rate for $N\geq2Kr$ and at, $M=\frac{\binom{C}{r}}{C}$, the improved S-LFR scheme is within a constant multiplicative factor from the optimal rate for $N\geq2K$.