Feature Selection Techniques for IoT Cybersecurity: A Review of Machine Learning-Based Intrusion Detection Approaches

Abstract

The rapid growth of Internet of Things (IoT) devices has introduced complex security challenges, necessitating efficient intrusion detection systems (IDS). This review explores the role of feature selection (FS) in enhancing the performance of machine learning-based IDS for IoT networks. Based on 20 peer-reviewed studies published between 2018 and 2025, FS methods are categorized into filter, wrapper, embedded, hybrid, and optimization-based approaches. The review evaluates these techniques using commonly adopted datasets such as NSL-KDD, CICIDS2017, BoT-IoT, and UNSW-NB15, and assesses their ability to address key IoT challenges, including limited resources, real-time processing, and data imbalance. Findings indicate a growing preference for hybrid and ensemble-based FS techniques due to their balance between accuracy and efficiency. However, limitations such as high computational costs and limited adaptability remain. This study offers a refined taxonomy, comparative analysis, and key recommendations for developing lightweight, adaptive, and explainable FS strategies suited for dynamic IoT security environments.