Enhancing Risk Management with Human Factors in Cybersecurity Using Behavioural Analysis and Machine Learning Technique (Published)
This study presents the development of an intelligent cybersecurity risk management system that leverages behavioural analytics and machine learning to detect threats and anomalous user activities in real time. The system was developed in the Extreme Programming (XP) methodology in certain important stages such as gathering of data, designing of the model, implementing it, and testing of the same. A deep learning model which was a hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model was involved to capture the spatial-temporal features of user behaviour logs and a Random Forest that acted as the final decision layer in anomaly classification. The process trained and assessed a complete set of information for nearly 411,000 records, consisting of the CERT Insider Threat Dataset v6.2, phishing email archives and simulated network/ system activity. The obtained results were shown to have good detection performance where the CNN-LSTM model had the highest mean accuracy 95.9%, precision 95.1%; recall 94.0%; and F1-score 94.5%. The Random Forest also increased the accuracy of classification. The real-time abilities and adaptive architecture of the system make it a feasible reality toward proactive and smart management of risks-related cybersecurity solutions in agile business environments.
Keywords: CNN-LSTM, Cybersecurity, Random Forest, Risk Management, anomaly detection, behavioural analytics