Game Theory based Consensus Learning in Unidirectional Rings

Abstract

This paper discusses the game theory based consensus learning in unidirectional rings. Distributed systemfindstheambiguousprocessorsorrationalagentsinthesamewayasgametheoryfindsambiguous players. The rational agents are those processors which deviate from the protocol and instead they follow their own path in order to fulfil their utility. In a bidirectional rings, the rational agents can be detected easily by each neighbourhood agents. But, in unidirectional case, the rational agents cannot be easily detected. This limits the honest agent to detect incorrect IDs sent by adversarial agents. So, our main focus is on the asynchronous unidirectional ring. We study the shortcomings of the protocol presentedin(Abrahametal.2013)and(Afeketal.2014).Inthispaper,wepresentanimprovedversion oftheprotocolbasedonGameTheorywhichismoreresilienttothepreviousone,byapplyinginvasion of adversarial processors and by proving it to be resilient in presence ofadversaries.