Spatio-Temporal Correlation based Anomaly Detection and Identification Method for IoT Sensors

Status monitor and anomaly detection of sensor nodes is critical for reliability of internet of things (IoT) system. Due to the complex causes of sensor anomalies, traditional methods that estimate only sensor state or environment state when detect anomalies cannot satisfy practical requirements. And existing methods often neglect spatial and temporal dependencies among IoT sensor observations. This paper defines faulty nodes and event nodes and proposes a novel anomaly detection and identification method implemented in two stages: 1) Anomaly detection stage by a customized composite distance metric and sensor clustering. 2) Anomaly source identification stage by fuzzy logic system based on spatio-temporal correlation. Experiments demonstrated that the proposed method can effectively detect anomaly sensors and capture spatio-temporal correlation to identify the anomaly source.