Abstract
Trust and reputation management emerges as a significant research trend, in term of soft security to tackle the security issues in computer networks. It is different from the traditional security mechanisms such as cryptography that is described as hard security. The basic idea is that every entity in the network, as an individual, can rate each other based on previous experiences. This rating on trust can assist other machines in deciding whether to collaborate with that machine in the future. Recently there has been a rapid increase in literature on trust and reputation management that mainly focuses on algorithmically modelling and evaluating the trust to effectively detect and avoid various malicious attacks. These trust algorithms can isolate the malicious entities from the local trust aspect. While the concept of trust in the computer network is derived from the sociology, and in sociology, it is defined as the belief that trustees will have a positive expectation of intention and behaviours. Moreover, the trustee at different positions will behave differently, such as at the Structural Hole or the position surrounded with Simmelian Ties. Do these position-based phenomena also exist in computer networks? In other words, in computer networks, is the location of a node will affect its behaviours, especially in the emerging peer-to-peer (P2P) network architecture?
Motivated by above research questions, in this thesis, we have focused on studying how the underlying network topological connectivity can affect the overlay trust behaviours from the global network perspective. This thesis has four main contributions. Firstly, we have revealed the underlying topology impact on the overlay trust behaviours in P2P networks. We have confirmed the correlations between the topological structures of Simmelian Ties and Structural Hole, and the node trustworthiness behaviours. Secondly, we have defined a new term of network trustworthiness to describe the trust level on a network topology. This is followed by introducing the Network Trustworthiness as a Service (NTaaS) concept, which can be adapted to accommodate the different levels of trust service demands from the users. Thirdly, we have proposed the $T$ value and Trustworthiness Tolerance Margin (MTT) based evaluation framework to evaluate the trustworthiness of the network topologies from the global aspect. Lastly, we have proposed a mathematical approach to optimise the network topology by adding a link in the most critical position so that the underlying network structures can best resist various unwanted behaviours and network failures.
