Optimal Voting Strategy against Random and Targeted Attacks

Abstract

Replication and value selection through voting are commonly used approaches to tolerating naturally caused failures. Without considering intentionally introduced failures, such as failures caused by attacks, having more replication or residency often makes the system more reliable. However, when both the reliability of individual replicas and the existence of attackers are taken into consideration, the number of replicas that participate in a voting process has significant impact on system reliability. In this paper, the authors study the problem of deciding the optimal number of participating voters that maximizes the reliability of voting results under two different types of attacks, i.e., random attack and targeted attack, and develop algorithms to find the optimal voting strategy. A set of experiments are performed to illustrate how the optimal voting strategy varies under different system settings and how the number of voting participants affects the system's reliability.