Date of Award

2023

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Department

Electrical & Electronic Engineering

First Advisor

Dr. Susan Rea

Second Advisor

Dr. Alan McGibney

Third Advisor

Dr. Ramona Marfievici

Abstract

In the last ten years, the Internet of Things (IoT) has experienced significant growth and has become a critical aspect of digital transformation across various industries such as healthcare, manufacturing, energy and society more broadly. IoT systems are now integrated across these verticals, creating ecosystems supporting data-driven decision-making and autonomous processes. These ecosystems consist of IoT network components, network services, and network participants such as organisations, consumers, governments, and businesses. They need to be built on the principles of the next-generation internet, which is viewed as being highly adaptive and resilient, trustworthy, sustainable, open and human-centric. The provision of Trust is a crucial component to align IoT ecosystems with these principles. Trust enables reliable data access, sharing, exchange, and processing, allowing for the provision of high-quality, context-aware intelligent services. However, managing trust in IoT ecosystems is challenging, considering the nature of the network i.e., resource constrained devices, the need for lightweight protocols, architectural limitations and the need to manage multiple stakeholders. Thus, the main objective of this research is to identify and explore these challenges in the context of creating a decentralised, highly adaptive, resilient and scalable IoT trust management framework. This dissertation proposes a trust framework that addresses the design and operational challenges of trust management within IoT ecosystems. The potential of Distributed Ledger Technology (DLT) as a mechanism to evaluate and maintain trust is critically evaluated and leveraged within this work to provide a technological grounding on which the framework is built. The framework consists of a trust architecture to enhance existing IoT-edge-cloud architectures through a novel DLT layer, which enables decentralised governance suitable for the multi-stakeholder settings. It supports the offloading of trust management functionalities, such as identity management, access control, and trust computation, from the edge and cloud while increasing traceability, accountability and provenance. Under the proposed trust framework, a robust, dynamic, multi- layer trust reputation model has been introduced to manage operational trust. The proposed model computes reputation algorithms on-chain in the DLT layer to support redundancy, resilience and trust. Moreover, it utilises Machine Learning (ML) algorithms within the reputation model’s core to automatically identify untrusted behaviours across IoT layers. This is a novel approach as it considers multi-layer IoT observations for computing trust. A trusted off-chain ML pipeline is proposed to facilitate the provisioning of ML models securely, which can be configurable for use in various domains. In order to demonstrate as well as evaluate the proposed framework, a proof- of-concept was implemented based on a technology stack that considered general IoT system requirements, to make the system widely applicable. The IoT network was implemented using the Routing Protocol for Low-Power and Lossy Networks (RPL), edge and cloud components leveraged Python-based libraries and the DLT layer is developed based on the Hyperledger Fabric open source framework. The untrusted behavioural detection ML models used as part of the framework achieved an overall accuracy of above 92%. Reputation computed based on these models clearly distinguished trusted and untrusted nodes, and 90% of the scenarios demonstrate convergence. The framework scalability evaluation shows that it can handle 1000 simultaneous trust update transactions in less than 10 seconds under the allocated hardware resources. This is a 100X improvement when compared to Ethereum-based trust computation models.

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Access Level

info:eu-repo/semantics/openAccess

Download

Share

COinS