A systematic review of metaheuristics-based and machine learning-driven intrusion detection systems in IoT

A team of researchers led by Mohammad Shamim Ahsan and Salekul Islam has published a comprehensive systematic review on arXiv, analyzing the integration of metaheuristic algorithms with machine learning (ML) for building Intrusion Detection Systems (IDS) in the Internet of Things (IoT). The paper, titled 'A systematic review of metaheuristics-based and machine learning-driven intrusion detection systems in IoT,' addresses a critical challenge: while ML improves threat detection, it demands excessive computing power ill-suited for the limited resources of IoT networks. The review finds that nature-inspired metaheuristic optimizers—algorithms modeled on animal behavior or natural phenomena—are being successfully applied to tasks like feature selection and hyperparameter tuning, boosting detection performance by up to 40% while slashing computational overhead.

The study's significant contribution is its discovery of hidden correlations between specific optimization techniques and ML models within state-of-the-art IoT-IDS architectures. It proposes a new taxonomy for existing systems and separately analyzes the effectiveness of metaheuristics in key applications. Furthermore, the researchers investigate critical issues related to this integration and conclude by discussing promising future algorithms and technologies, such as those suited for decentralized computing trends, that could further enhance IoT security efficiency. This work serves as a vital roadmap for security developers aiming to build robust, lightweight defensive AI for the billions of connected devices coming online.