In recent years, there has been a significant increase in insider threats within organisations and these have caused massive losses and damages. Due to the fact that email communications are a crucial part of the modern-day working environment, many insider threats exist within organisations’ email infrastructure. It is a well-known fact that employees not only dispatch ‘business-as-usual’ emails, but also emails that are completely unrelated to company business, perhaps even involving malicious activity and unethical behaviour. Such insider threat activities are mostly caused by employees who have legitimate access to their organisation’s resources, servers, and non-public data. However, these same employees abuse their privileges for personal gain or even to inflict malicious damage on the employer. The problem is that the high volume and velocity of email communication make it virtually impossible to minimise the risk of insider threat activities, by using techniques such as filtering and rule-based systems. The research presented in this dissertation suggests strategies to minimise the risk of insider threat via email systems by employing a machine-learning-based approach. This is done by studying and creating categories of malicious behaviours posed by insiders, and mapping these to phrases that would appear in email communications. Furthermore, a large email dataset is classified according to behavioural characteristics of employees. Machine learning algorithms are employed to identify commonly occurring insider threats and to group the occurrences according to insider threat classifications. / Dissertation (MSc (Computer Science))--University of Pretoria, 2020. / Computer Science / MSc (Computer Science) / Unrestricted

Insider threat is one of the greatest concerns for information security that could cause more significant financial losses and damages than any other attack. However, implementing an efficient detection system is a very challenging task. It has long been recognized that solutions to insider threats are mainly user-centric and several psychological and psychosocial models have been proposed. A user's psychophysiological behavior measures can provide an excellent source of information for detecting user's malicious behaviors and mitigating insider threats. In this dissertation, we propose a multi-modal framework based on the user's psychophysiological measures and computer-based behaviors to distinguish between a user's behaviors during regular activities versus malicious activities. We utilize several psychophysiological measures such as electroencephalogram (EEG), electrocardiogram (ECG), and eye movement and pupil behaviors along with the computer-based behaviors such as the mouse movement dynamics, and keystrokes dynamics to build our framework for detecting malicious insiders. We conduct human subject experiments to capture the psychophysiological measures and the computer-based behaviors for a group of participants while performing several computer-based activities in different scenarios. We analyze the behavioral measures, extract useful features, and evaluate their capability in detecting insider threats. We investigate each measure separately, then we use data fusion techniques to build two modules and a comprehensive multi-modal framework. The first module combines the synchronized EEG and ECG psychophysiological measures, and the second module combines the eye movement and pupil behaviors with the computer-based behaviors to detect the malicious insiders. The multi-modal framework utilizes all the measures and behaviors in one model to achieve better detection accuracy. Our findings demonstrate that psychophysiological measures can reveal valuable knowledge about a user's malicious intent and can be used as an effective indicator in designing insider threat monitoring and detection frameworks. Our work lays out the necessary foundation to establish a new generation of insider threat detection and mitigation mechanisms that are based on a user's involuntary behaviors, such as psychophysiological measures, and learn from the real-time data to determine whether a user is malicious.