MACHINE LEARNING-BASED ENSEMBLE APPROACH FOR MALWARE DETECTION IN CLOUD SYSTEMS

Abstract

Malware poses a significant security threat to cloud computing environments, particularly as emerging and obfuscated threats increasingly challenge conventional detection techniques. This study proposes SpectraStack-Net, a stacking ensemble framework for malware detection that integrates Random Forest, Extra Trees, and XGBoost as base classifiers with Logistic Regression as a meta-classifier. The proposed framework incorporates data preprocessing, mutual-information-based feature selection, and SMOTE-based class balancing to improve classification performance. Experiments were conducted on a cleaned dataset of 216,352 unique samples, using a stratified 80:20 train-test split. The proposed ensemble achieved 99.64% accuracy, 99.60% precision, 99.38% recall, 99.49% F1-score, 99.22% MCC, and 99.99% AUC, outperforming the individual classifiers across the evaluated metrics. The results demonstrate that combining heterogeneous machine learning models through stacking provides a highly accurate and reliable approach for malware detection in cloud environments.

Keywords: Malware Detection; Cloud Security; Stacking Ensemble Learning; Machine Learning; Cybersecurity.