(A) study on fast and reliable distributed information retrieval and consensus

Abstract

This dissertation studies communication schemes for fast and secure retrieval and consensus of distributed information. First of all, a novel capacity achieving scheme for the private information retrieval is proposed. Private information retrieval is a communication scheme that aims to retrieve a desired information without disclosing the identity of the desired information. Differentiated from the traditional capacity achieving scheme for the private information retrieval, the proposed scheme is based on stochastic and asymptotic approach. Compared to the conventional scheme, the proposed scheme offers less complex solution of designing answering strings that ensures privacy and achieves the capacity at the same time. In addition, fundamental limits of private information retrieval aided with coded side information are discovered. Specifically, the cost for retrieving the desired information is examined by the normalized download cost, which counts the number of bit downloaded for a single desired bit with securing the privacy. The achievability of the fundamental limits are also shown via stochastic and asymptotic approach as done for the private information retrieval without side information. Next, a fast and scalable distributed consensus protocols are studied in the perspective of wireless communication. The traditional consensus protocols were usually ill-suited for the network that is composed of many users. In this context, this paper proposes a fast consensus protocol based on a co-design of communication and consensus process. Its feasibility and effectiveness is shown both analytically and numerically, Moreover, its applicability is underpinned by the study on of the fast consensus protocol applied dynamic spectrum access technique called `Consensus-Before-Talk'.