Analytical metric weight generation for multi-domain trust in autonomous underwater MANETs

Trust Management Frameworks (TMFs) are being used to improve the efficiency, security, and reliability of decentralized and distributed autonomous MANETs using metrics garnered from the communications activities of nodes within the networks. However, these do not perform well in sparse / harsh environments such as those found in Underwater Acoustic Networks (UANs) [1]. As node capabilities increase, the physical motion of nodes represent an additional domain of knowledge about the operations and behaviours of the network. In this paper we present a Machine Learning supported methodology for optimising metric weight vector generation, using metrics from both physical and communications domains to detect and identify a range of misbehaviours, demonstrating that by utilising information from multiple domains, trust assessment can be more sensitive and accurate than in single-domain (communications) assessment.