A framework for anomaly classification in Industrial Internet of Things systems

Abstract

ABSTRACT: Introducing the Industrial Internet of Things (IIoT) into traditional industrial processes has marked a new era of enhanced connectivity and productivity. By integrating advanced sensors, communication technologies, and data analysis, IIoT enables real-time monitoring, proactive maintenance, and increased operational efficiency. However, this increased complexity and interconnectivity also introduce new challenges in maintaining system dependability and safety. Considering these issues, this work presents an IIoT Anomaly Classification Framework designed to detect and categorize anomalies such as failures and attacks. The research addresses the critical need for robust anomaly detection and classification in IIoT systems by providing a comprehensive and scalable solution adaptable to various industrial contexts. The framework comprises two main components: an anomaly detection model and an anomaly classification model. The anomaly detection model operates unsupervised, continuously monitoring system data to identify deviations from normal behavior patterns. At the same time, the anomaly classification model categorizes these anomalies based on historical data using machine learning algorithms. The proposed framework has been tested in a realistic IIoT environment, demonstrating its effectiveness and practicality. During the cross-validation process, a precision of 0.95, recall of 0.88, and F1-score equal to 0.91 were obtained. This research contributes significantly to IIoT, offering a valuable tool for improving industrial operations and laying the groundwork for future anomaly classification and system resilience advancements.