The rapid expansion of Internet of Things (IoT) devices has significantly heightened the risk of cyber-attacks, necessitating robust security measures such as Intrusion Detection Systems (IDS). Traditional IDS approaches often struggle to process the vast amounts of data generated by IoT networks. Recent advancements in deep learning, particularly Convolutional Neural Networks (CNNs), have demonstrated potential in enhancing security by automatically detecting complex patterns in data. This study aims to explore and analyze CNN-based and hybrid deep learning methods for IDS in IoT networks. Design/Methodology/Approach: This paper conducts a comprehensive review of IDS techniques for IoT networks, focusing on deep learning-based approaches published between 2017 and 2025. The analysis includes methods such as edge computing, transfer learning, lightweight models, and federated learning to improve detection accuracy and efficiency while addressing the resource constraints of IoT devices. Findings/Result: The study highlights that deep learning-based IDS can effectively detect both known and unknown threats with high accuracy and low false-positive rates. However, challenges such as high computational costs, interpretability of models, and real-time processing limitations remain. The review identifies key factors influencing IDS performance and proposes potential improvements for future research.
Shaikh et al. (Wed,) studied this question.