Enhancing Cyber threat modelling by applying innovative machine learning approaches

The objective of my research is to develop a Threat Intelligence Framework that augments threat data, to provide value to securing information and assets on a strategic, and operational level. The framework will provide actionable threat intelligence using a learn, adapt, suggest and action approach. It will focus on a four-phased approach that collects, analyses, identifies and predicts threats using existing machine learning techniques identified through my systematic and critical review process.
I will research and review the most effective approach to extract data from threat repositories. The collection phase will then provide input to the next phase that will contextualise the acquired threat data based on asset features and configuration. This phase will then identify the supporting assets and their relevant threats, vulnerabilities and potential exploits. The third phase of the system will be to analyse the data and model collective existing known/unknown threats, and the resulting likelihood of a compromise. The output of this assessment phase will provide the features used in the threat prediction phase, that then iteratively feeds back into the previous (third) phase.
The proposed research will compare machine learning algorithms and identify the best combined Machine Learning Algorithms in the 4 Phases of the framework. The asset risks identified from vulnerabilities, threat intelligence and models, will be part of a feedback loop, providing itself with training data. The iterative system will continue to analyse and put into context asset/threat/vulnerability data and attempt to make point-in-time predictions on known, unknown and emerging threats.