Evaluation of Machine Learning Implementation for Network Intrusion Detection in Distributed IoT Systems

The rapid expansion of Internet of Things (IoT) ecosystems has significantly increased cybersecurity risks due to device heterogeneity, limited computational resources, and distributed network architectures. Traditional security mechanisms are insufficient to address evolving threats such as Distributed Denial of Service (DDoS), botnets, and zero-day attacks. This study aims to evaluate the implementation of machine learning (ML) algorithms for network intrusion detection in distributed IoT systems by examining accuracy, efficiency, and scalability. The research employs a qualitative literature review approach, systematically analyzing reputable journal articles and conference papers related to IoT security, Intrusion Detection Systems (IDS), and machine learning applications. Data were collected through identification, selection, and thematic synthesis of relevant studies, focusing on algorithm types, evaluation metrics, architectural models, and implementation challenges. The results indicate that deep learning models provide superior accuracy in detecting complex and evolving attacks, while traditional machine learning algorithms offer better computational efficiency for edge deployment. Furthermore, distributed and federated learning architectures enhance scalability and reduce communication overhead. A hybrid hierarchical approach integrating edge, fog, and cloud layers is identified as the most effective solution.