Enhancing Insider Threat Detection with Machine Learning Techniques

Insider threats pose significant risks to organizations by compromising sensitive data and resources. Detecting these threats effectively necessitates robust machine learning (ML) techniques capable of managing complex and imbalanced datasets. This paper explores the performance of various ML models, including Logistic Regression, Decision Trees, Random Forest, SVM, KNN, Naive Bayes, Adaboost, and XGBoost, using the widely recognized CERT dataset. By addressing data imbalance challenges through techniques such as SMOTE, the importance of a balanced dataset is highlighted. The results demonstrated that Random Forest and Adaboost achieved the highest accuracy of 97.5<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, underscoring their effectiveness in insider threat detection. This research contributes to enhancing insider threat detection methodologies and provides a structured analysis of model performance, paving the way for more reliable organizational security strategies.