In today’s digital landscape, safeguarding sensitive data, especially in critical sectors like finance, is vital. This study aims to enhance Host Intrusion Detection Systems (HIDS) at Ethiopia’s Ministry of Finance using machine learning and deep learning techniques. It compares different methods of supervised and unsupervised learning to identify the most effective method for detecting threats with the Ministry’s datasets. Various algorithms, including Random Forest and XGBoost for feature selection, and techniques like SMOTE for imbalanced classes, were employed. In supervised learning, methods like Random Forest, Decision Tree, KNN, XGBoost, Nave Bayes, Logistic Regression and SVM were evaluated. In unsupervised learning, techniques like Isolated Forest, LOF, One-class and SVM were explored. The analysis revealed notable performance, with supervised models achieving perfect accuracy 100%, particularly SVM. The Isolated Forest algorithm stood out in unsupervised learning with a 99% accuracy. Deep learning, especially the denoising autoencoder, showed promising results, achieving a lower Reconstruction error of 0.17.