Use this URL to cite or link to this record in EThOS:
Title: An investigation into trust and reputation frameworks for autonomous underwater vehicles
Author: Bolster, Andrew
Awarding Body: University of Liverpool
Current Institution: University of Liverpool
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
As Autonomous Underwater Vehicles (AUVs) become more technically capable and economically feasible, they are being increasingly used in a great many areas of defence, commercial and environmental applications. These applications are tending towards using independent, autonomous, ad-hoc, collaborative behaviour of teams or fleets of these AUV platforms. This convergence of research experiences in the Underwater Acoustic Network (UAN) and Mobile Ad-hoc Network (MANET) fields, along with the increasing Level of Automation (LOA) of such platforms, creates unique challenges to secure the operation and communication of these networks. The question of security and reliability of operation in networked systems has usually been resolved by having a centralised coordinating agent to manage shared secrets and monitor for misbehaviour. However, in the sparse, noisy and constrained communications environment of UANs, the communications overheads and single-point-of-failure risk of this model is challenged (particularly when faced with capable attackers). As such, more lightweight, distributed, experience based systems of “Trust” have been proposed to dynamically model and evaluate the “trustworthiness” of nodes within a MANET across the network to prevent or isolate the impact of malicious, selfish, or faulty misbehaviour. Previously, these models have monitored actions purely within the communications domain. Moreover, the vast majority rely on only one type of observation (metric) to evaluate trust; successful packet forwarding. In these cases, motivated actors may use this limited scope of observation to either perform unfairly without repercussions in other domains/metrics, or to make another, fair, node appear to be operating unfairly. This thesis is primarily concerned with the use of terrestrial-MANET trust frameworks to the UAN space. Considering the massive theoretical and practical difference in the communications environment, these frameworks must be reassessed for suitability to the marine realm. We find that current single-metric Trust Management Frameworks (TMFs) do not perform well in a best-case scaling of the marine network, due to sparse and noisy observation metrics, and while basic multi-metric communications-only frameworks perform better than their single-metric forms, this performance is still not at a reliable level. We propose, demonstrate (through simulation) and integrate the use of physical observational metrics for trust assessment, in tandem with metrics from the communications realm, improving the safety, security, reliability and integrity of autonomous UANs. Three main novelties are demonstrated in this work: Trust evaluation using metrics from the physical domain (movement/distribution/etc.), demonstration of the failings of Communications-based Trust evaluation in sparse, noisy, delayful and non-linear UAN environments, and the deployment of trust assessment across multiple domains, e.g. the physical and communications domains. The latter contribution includes the generation and optimisation of cross-domain metric composition or“synthetic domains” as a performance improvement method.
Supervisor: Marshall, A. ; Maskell, S. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral