Beyond System Logs: Early Detection of Potential Insider Threats Using Psychometric HR Indicators and Ensemble Machine Learning

Abstract – The paradigm of cybersecurity is shifting from purely technical defenses to human-centric approaches. While traditional insider threat detection relies on reactive analysis of system logs, this study proposes a preventive framework by analyzing the root causes of malicious behavior: employee disgruntlement and cognitive fatigue. Utilizing a secondary dataset of 1,000 employee records, we constructed a "Motive-Opportunity" risk model based on psychometric and operational HR metrics. By employing an Ensemble Machine Learning approach (Random Forest), we successfully differentiated high-risk profiles from normal workforce behavior. The model achieved a classification accuracy of 93.6%, with a precision of 1.0 for high-risk identification. The results empirically confirm that low job satisfaction combined with excessive overtime are the strongest predictors of security vulnerability. This study bridges the gap between Human Resources and Information Security, offering a proactive methodology for organizations to mitigate insider threats before digital violations occur.

Keywords – Insider Threat, HR Analytics, Cybersecurity, Machine Learning, Random Forest.