A Robust Feature Integration for Multiclass Metamorphic Malware Detection in IoT Network

With the increase in the use of Internet of Things (IoT) services and applications, the escalating prevalence of metamorphic malware poses a significant challenge. Characterized by their ability to dynamically modify their code to evade detection, these advanced malware variants significantly compromise the security of IoT networks. This paper presents an approach for multiclass metamorphic malware detection in IoT networks, emphasizing the integration of diverse features by employing Convolutional Neural Networks (CNN) for intricate feature extraction, Principal Component Analysis (PCA) for eliminating multicollinearity between the features, and Random Forest (RF) for robust classification. Our proposed model demonstrates exceptional performance with macro-accuracy, macroprecision, macro-recall, and macro-F1 score of 97.44%, and a distinctive ROC-AUC score of 99.87%.