Work

Understanding the influence of behaviour monitoring tools towards mitigating insider threats in hybrid working environments through behavioural analytics

Year: 2025
Type: dissertation
Abstract: Hybrid working arrangements would pose a major challenge to the insider threat detection tools since the traditional monitoring tools are not very effective in detecting malicious activities in work environments that are dispersed. In 2023, organisations have suffered severe consequences due to insider threats, 45% led to data loss and 43% to brand damage, but less than one third of the organisations perceived that they were well prepared to counter threats. The study aimed to assess the efficiency of machine learning models (Random Forest, Stacking Classifier, and XGBoost) enhanced by behavioural analytics to detect insider threats and compare their results in relation to improving the model of child enhancement accuracy and reducing false positives in the hybrid workplaces. The research employed a positivist philosophy and the deductive approach based on the Kaggle Insider Threat dataset. Preprocessing of data involved the management of missing data, outliers, class balancing using SMOTE, and multicollinearity filtering. The accuracy, precision, recall, F1-score, and AUC were used to measure, train, and evaluate three ensemble models. XGBoost was the most suitable model, with 99.20% accuracy, 100% recall, and 98.57% F1-score, followed by Random Forest (99.07% accuracy, 97.58% F1-score) and Stacking Classifier (98.67% accuracy, 97.58% F1-score). The results show that the use of ML to enhance behavioural monitoring performs well in detecting insider threats in the hybrid environment, as it is more effective than the available methods reported in the literature, and provides support to the use of ensemble methodologies in organisational cybersecurity resilience.
Source: NORMA
Language: English
Open Access status: green
OpenAlex ID: W7210051119
Cites: 17
Cited by:
Related to: 10