An Effective Trust System for Outlier Detection in MANETs using Machine Learning

Abstract

The applications of MANETs in critical scenarios like battlefield message transmission and emergency relief communication requires the adoption of security mechanisms which depend upon cooperation among nodes. If any adversary manages to enter such type of networks, it can easily introduce vulnerabilities to initiate malicious activities, such as dropping of packets, modification etc. One possible way to protect against this type of outlier behavior is to create a trust environment. In general, Machine Learning is used to forecast future behavior of the nodes by looking into their previous history of interactions. In this paper, a trust system based upon Machine Learning (ML) is proposed which is able to protect against various patterns of attacks. The results show that our trust based system significantly identifies outlier nodes even when the proportion of malicious nodes in the network is high.