Deep Learning-Based Intrusion Detection Systems for IoT Networks: A Systematic Literature Review and Comparative Analysis

Abstract

The new generations of communication networks are demanding intrusion detection systems capable of addressing sophisticated cyber threats. This systematic literature review examines deep learning-based intrusion detection systems for IoT networks through rigorous analysis, adhering to PRISMA 2020 guidelines. We synthesize findings from studies to address five research questions covering IoT security challenges, architectural approaches, performance characteristics, emerging research directions, and providing taxonomy of deep learning architectures, strategies and applications. Our analysis indicates that hybrid deep learning architectures report higher metrics than single-model approaches in evaluated scenarios. Critical research gaps emerge across multiple dimensions, such as edge deployment limited resources, lack of realistic IoT-specific datasets and absence of explainable AI mechanisms in current solutions. This synthesis provides insights for advancing IoT security.