Bayesian Network Model for a Zimbabwean Cybersecurity System
Gabriel Kabanda
Abstract
The purpose of this research was to develop a structure for a network intrusion detection and prevention system based on the Bayesian Network for use in Cybersecurity. The phenomenal growth in the use of internet-based technologies has resulted in complexities in cybersecurity subjecting organizations to cyberattacks. What is required is a network intrusion detection and prevention system based on the Bayesian Network structure for use in Cybersecurity. Bayesian Networks (BNs) are defined as graphical probabilistic models for multivariate analysis and are directed acyclic graphs that have an associated probability distribution function. The research determined the cybersecurity framework appropriate for a developing nation; evaluated network detection and prevention systems that use Artificial Intelligence paradigms such as finite automata, neural networks, genetic algorithms, fuzzy logic, support-vector machines or diverse data-mining-based approaches; analysed Bayesian Networks that can be represented as graphical models and are directional to represent cause-effect relationships; and developed a Bayesian Network model that can handle complexity in cybersecurity. The theoretical framework on Bayesian Networks was largely informed by the NIST Cybersecurity Framework, General deterrence theory, Game theory, Complexity theory and data mining techniques. The Pragmatism paradigm used in this research, as a philosophy is intricately related to the Mixed Method Research (MMR). A mixed method approach was used in this research, which is largely quantitative with the research design being a survey and an experiment, but supported by qualitative approaches where Focus Group discussions were held. The performance of Support Vector Machines, Artificial Neural Network, K-Nearest Neighbour, Naive-Bayes and Decision Tree Algorithms was discussed. Alternative improved solutions discussed include the use of machine learning algorithms specifically Artificial Neural Networks (ANN), Decision Tree C4.5, Random Forests and Support Vector Machines (SVM).