In the fast-evolving landscape of cybersecurity, the reliable detection of malicious software is a formidable challenge that requires robust and dependable solutions. This study introduces a novel malware detection framework utilizing an uncertainty-aware ensemble model, designed to enhance prediction precision and reliability by integrating ensemble learning (EL) with uncertainty quantification. Our framework employs a stacked ensemble model where an Artificial Neural Network (ANN) acts as the meta classifier, supported by AdaBoost (AB), Logistic Regression (LR), and Support Vector Machine (SVM) as base classifiers. By leveraging the diverse strengths of each base classifier, this ensemble model significantly improves malware detection’s robustness and adaptability. To address the inherent unpredictability in malware detection, the framework incorporates Monte Carlo dropout (MCD), enabling systematic uncertainty analysis and quantification of prediction confidence. Extensive experiments conducted on a publicly available dataset demonstrate the framework’s effectiveness, achieving a remarkable accuracy of 99.98%.

Moreover, the uncertainty analysis underscores the model’s capability to generate highly certain predictions, underscoring its suitability for real-world cybersecurity applications.