Enhancing Insider Threat Detection with Machine Learning Techniques

Abstract

The compromise of confidential information and company assets by insiders is a major concern for
every business. For these dangers to be detected, strong ML methods are required, ones that can handle skewed and
complicated datasets. This research examines the efficacy of several machine learning models on the renowned
CERT dataset. These models include Logistic Regression, Decision Trees, Random Forest, SVM, KNN, Naïve
Bayes, Adaboost, and XGBoost. Methods like SMOTE bring attention to the significance of a balanced dataset by
fixing problems caused by data imbalance. With an accuracy of 97.5%, Random Forest and Adaboost proved to be
the most successful in detecting insider threats, according to the data. Improved organizational security strategies are
possible because to this research's contributions to insider threat detection approaches and structured examination of
model performance.