Distributed binary majority voting via exponential distribution

Abstract

In the binary majority voting problem, each node initially chooses between two alternative choices. The goal is to design a distributed algorithm that informs nodes which choice is in majority. In this study, the authors formulate this problem as a hypothesis testing problem and propose fixed-size and sequential solutions using classical and Bayesian approaches. In the sequential version, the proposed mechanism enables nodes to test which choice is in majority, successively in time. Hence, termination of the algorithm is embedded within it, contrary to the existing approaches which require a monitoring algorithm to indicate the termination. This property makes the algorithm more efficient in terms of message complexity. Furthermore, the authors show that the proposed solution is resilient to Byzantine attacks if network connectivity is F + 1 in the presence of F adversarial nodes. Thus, the proposed algorithm is more robust compared with the previous works which are vulnerable to the existence of adversarial nodes.

Distributed Voting in Beep Model

“…Some of these algorithms have been proposed for the binary case. Binary voting using randomized gossip algorithm [43], average consensus [44], two-sample voting [45], and via exponential distribution [46] are the examples of binary voting algorithms. There are also some methods for multi-choice voting, such as voting using pairwise asynchronous graph automata [47] and union and intersection operations [48].…”