An Improved Model for Fault Detection and Isolation in an IoT-Based Network

Abstract

The rapid growth of Internet of Things (IoT) deployments has increased the need for dependable mechanisms for detecting and isolating abnormal network behaviour. Packet loss, latency, congestion, and intermittent connectivity can degrade data integrity, service continuity, and operational reliability, particularly in distributed and resource-constrained environments. This study developed an enhanced Fault Detection and Isolation System (IoTFDS) that combines supervised Random Forest classification with unsupervised Isolation Forest anomaly isolation. Network traffic data were cleaned, normalized and divided into training and testing subsets before model development. Random Forest was used to classify network states as normal or faulty, while Isolation Forest was applied to abnormal records to identify anomalous patterns and support fault localization. An intermediate evaluation produced 69.4% accuracy with zero true positives, demonstrating the importance of model refinement and class-sensitive evaluation. The final system also provided an operational interface for network monitoring, fault detection, anomaly isolation and result visualization. The findings indicate that the hybrid approach can provide accurate classification and complementary anomaly isolation for IoT network monitoring, although validation on live IoT traffic and clearer documentation of dataset provenance remain necessary