To Spy the Lie. Detecting the Insider Threat of Espionage

Abstract: Acts committed by insiders have risen during past years, and there is a need for a better understanding of how preventive measures can be used, not just remedial action after the fact. The current narrative in research when discussing espionage was motive; why someone committed espionage. The aim of this study was to create a theoretical model of a ‘risk individual’ and, with the use of the model, techniques for personality assessment and text analysis, develop an artefact, a self-assessment test, that could be used to assess if a person had a higher risk to commit the act of espionage. Design Science research was chosen as a main methodological approach with supporting methods throughout. A survey was chosen to collect the data and the data was analyzed quantitatively. The artefact is partly based on selfassessment questionnaires and partly on themes identified as necessary when a governmental agency conducted personal security interviews for potential new hires. In order to achieve the research goal, data from 52 individuals were collected and analyzed using various quantitative methods. When applying internal reliability testing to the risk factors proposed by the theoretical model, seven out of the eight factors had good reliability. One factor, stress, performed poorly. This was probably due to the width of the questions asked, from personal to professional stress. This resulted in stress being removed from further testing. The remaining seven factors correlated with each other, apart from one, entitlement. This risk factor correlated with ethical flexibility but not the other six risk factors. In order to test how well the Big Five correlated with risk, the mean of a risk individual was calculated and compared with the five factors of OCEAS. The five factors all correlated negatively with risk, with agreeableness having the highest negative correlation and extroversion having the lowest. Differences could be seen when comparing the ten participants with the highest mean risk score to the ten with the lowest for both the Big Five and the risk factors in the theoretical model. The differences for the Big Five were lower than those for the theoretical model, i.e., both Big Five and the theoretical model work as sorting out higher-risk individuals. However, they worked better together and provided a more profound picture than using just one or the other. The open-text questions were analyzed with the help of wordlists to calculate how the participants used different types of pronouns when writing. One wordlist provided potentially interesting results (the word list for ‘I’), while the others did not.