Feature selection with machine learning of personality differences and insider theft of information

Abstract

Current strategy to detect insider threat has been heavily from technical strategy relying on analysis on large amount of data and complex data analysis. It has been proposed that putting psychological profiling or social engineering aspect into the system to detect insider threats several advantages are to be expected. However, in current literature of information technology only a number of research has been done to study human psychological behaviour perspective to support decision in detecting insider threat within enterprises. This paper aims to perform feature selection with machine learning to investigate personality differences and insider theft of information. The results from this study can also serve as a building block for future studies to develop an assessment tool that can capture relevant behavioural data in real time to help predict insider threats in organisations or enterprises. Data collection is done by conducting a deceptive experiment where participants invited to partake in a technical task, in this case a programming challenge in Matlab. Personality information is acquired from questionnaires administered post-experiment. Feature selection explored for data analysis is mainly a filter method with different evaluation measures. Two best features are extracted from each evaluation measures and discriminative power of the two features is investigated.