A New Hybrid Ensemble Learning-Based Malware Detection Technique

One of the most challenging difficulties that businesses and institutions face is information security. In recent years, the frequency and magnitude of cybercrime have increased, with innovative ways to access, manipulate, and delete data or render information systems disabled, appearing on a daily basis. Malware is one type of intrusion into information systems that process confidential data. For a long time, malware has posed a serious threat to computer system security. A hacker injects malware into a computer system, acquiring complete or partial accessibility to the system’s critical data. Conventional detection methodologies based on dynamic as well as static analysis have limited effectiveness due to the rapid advancement of anti-detection technologies. AI-based malware detection has grown in popularity in recent years because of its superior prediction ability. As a result, this research presents a hybrid ensemble classification-based approach for information security and malware detection. A stacked ensemble of 5 homogenous machine learning algorithms performs the first stage classification, where each model is itself ensembled 5 times individually, while in the the final stage, classification is carried out by assembling these 5 ensembled models together, resulting in a meta-learner. Individually, the following machine learning algorithms are utilized for baseline comparison: K-Nearest Neighbors, Support Vector Machine (SVM), Logistic-Regression, Naive Bayes and Decision Tree. Further, Malware Classification using PE headers (ClaMP) dataset was tested, and the findings are reported. It has been found that the proposed hybrid ensemble model achieves the optimum performance.