A cluster-based Bayesian approach to reliable route discovery in large-scale mobile ad hoc networks

Abstract

The application of MANETs have become a central point of focus in the development of wireless networks. However, MANETs still face the challenge of reliable route discovery in large-scale networks under adverse node failure conditions. To address this issue, a novel adaptation of the AODV protocol called the Na¨ıve Bayes Classifier Routing (NBCR) pro-tocol is presented. Furthermore, an extension of this protocol via a clustering overlay is also presented, and respectively called the Cluster-based Na¨ıve Bayes Classifier Routing (CB-NBCR) protocol. Using Bayes' theorem and a classifier probability framework, pos-terior route trust probabilities are calculated from prior evidence. These probabilities are used to select routes that would improve network performance on large-scale networks in the presence of dysfunctional/selfish nodes. The clustering overlay additionally attempts to improve network performance by reducing large ad hoc networks to more manageable topologies.