Analysis of binary voting algorithms for use in fault-tolerant and secure computing

Abstract

We examine three binary voting algorithms used with computer replication for fault tolerance and separately observe the resultant reliability and security. We offer insights to answer the question: Can a voting algorithm provide a system with both security and reliability? We show that while random dictator (i.e., randomly choosing one of the replicas) provides good security and majority rule yields good fault tolerance neither is effective in both. We present the random troika (a subset of 3 replicas) as an effective combination of fault-tolerant and secure computing.