Security, Secrecy, and Privacy in Hotplug and Multi-Access Coded Caching Problems

Abstract

With the rapid growth in data consumption, efficient content delivery has become a key challenge in modern communication networks. Coded caching offers a powerful solution by leveraging storage at the users to reduce peak traffic loads through coded multicasting. This technique was first introduced by Maddah-Ali and Niesen (MAN) in their work ''Fundamental Limits of Caching'' in IEEE Transactions on Information Theory, 2014. The network model consists of a server with $N$ files connected to $K$ users through an error-free shared link, where each user has a dedicated cache. The system operates in two phases: the placement phase, during which each user's cache is populated up to its capacity, and the delivery phase, where users declare their demands and the server responds accordingly. During delivery, the server leverages the cached content to minimise network traffic. In addition to the dedicated cache setup considered by MAN, this thesis also addresses coded caching problems in multi-access networks, where each user accesses more than one cache. New coded caching schemes are proposed for different system models with additional constraints. First, we consider a model with $H$ servers storing an MDS-coded file library, where each user accesses a unique set of $r$ out of $C$ caches, as in the combinatorial topology and retrieves a linear function of the files. Users must be able to recover their demands from any $L$ out of $H$ servers, while ensuring content security against an eavesdropper and demand privacy against non-colluding users and any colluding set of servers. Two schemes are proposed: one satisfying both security and privacy conditions, and another satisfying only the security requirement, both offering order-optimal performance in the low memory regime. These schemes also generalize prior work. Next, we consider a setup with $C$ multi-access caches and $K$ private caches, where each user accesses one private cache and the multi-access caches are accessed as in the combinatorial topology. A new scheme is proposed that achieves linear function retrieval, content security, and demand privacy against colluding users. We derive a cut-set based lower bound and provide some optimality results. The proposed scheme is extended to a more general setup where different users are connected to different numbers of multi-access caches, and multiple users are connected to the same subset of multi-access caches. Further, this thesis addresses the multi-access combinatorial topology with secrecy and demand privacy constraints. We propose a scheme satisfying both conditions and derive a lower bound based on cut-set arguments. The scheme is proven to be optimal under certain conditions. Finally, we focus on the hotplug coded caching model, where some users are offline during the delivery phase. A new hotplug scheme based on existing Hotplug Placement Delivery Arrays (HpPDAs) is proposed. The performance comparison of the proposed scheme with the existing schemes is presented. The proposed scheme outperforms some of the known schemes in the literature. Then, secrecy is incorporated into the hotplug coded caching setup. Two secretive schemes are proposed for the two known classes of HpPDAs. Numerical comparisons show that the proposed schemes outperform a baseline secretive scheme adapted from the classical coded caching setup in specific memory regions.