Abstract
The 21st century introduced a rapid increase in technological developments that change the way people go about their everyday lives by creating innovative new methods that replace existing outdated processes. Many of these processes relate to the way in which people choose to communicate and transfer information.
The digitisation of information provides easy access to a broad range of knowledge areas, however, it also presents the potential for exposure of personal and sensitive information through security vulnerabilities and malicious software. Biologically-inspired computation systems, known as artificial immune systems (AIS), have been applied to a broad range of domains to provide innovative solutions and enhancements to complex problems, for example, information security.
The application of an AIS to anomaly detection has proven to be a viable solution by many published research papers, however, the generation of effective detectors is constantly highlighted as an issue for improvement. The agent-orientated software development paradigm presents a potential solution to this issue through autonomous software objects that are capable of modelling complex structures such as the vertebrate immune system.
This dissertation introduces the MANS (Multi-Agent Negative Selection) model for the creation of an AIS which uses the biologically-inspired negative selection theory. The MANS model presents a multi-agent system approach to mimicking the behaviour of lymphocyte creation, training and application to the classification of elements as known or unknown, while providing self-tolerance.
Additionally, the MANS model contributes towards exploratory research and the creation of the multi-agent MANS prototype system. The prototype system provides a modification of the negative selection algorithm applied to character string file hashes and real-valued shapes within a high dimensional...
M.Sc. (Informatics)