Recent Advances and Challenges In Malware Detection For Internet of Things Systems: A Comprehensive Review

Abstract

The Internet of Things (IoT) ecosystem faces escalating security threats from sophisticated malware targeting diverse, resource-constrained devices. Despite advances, effective detection remains challenging due to evolving malware behaviors and heterogeneous environments.  This review presents a comprehensive survey of over forty-three studies from 2018 to 2025, analyzing machine learning, deep learning, hybrid, and non-AI-based malware detection techniques. Our review reveals that deep learning and hybrid models generally outperform traditional methods by capturing complex behavioral patterns, yet issues like limited dataset diversity, computational demands, and explainability persist. We identify critical research directions including lightweight edge-compatible models, federated learning, multimodal feature fusion, and explainable AI integration. These insights provide a structured understanding of current approaches and guide the development of scalable, robust, and interpretable IoT malware detection systems, advancing cybersecurity in increasingly connected environments.