Anomaly Detection Using Spatio-Temporal Correlation and Information Entropy in Wireless Sensor Networks

Wireless Sensor Network (WSN) has already been deployed widely in industrial, agriculture and many other fields. Meaningful information extracted from the data collected by WSNs can assist people in making control and decisions. Anomaly detection of sensor data is critical to ensure the data quality and the reliability of WSNs. In recent years, distance-based anomaly detection methods have attracted the attention of many experts. Though these methods can achieve a good detection effect, the exploitation of the spatio-temporal correlation of sensor data is not sufficient. Therefore, we propose an anomaly detection method (STCIE) using the spatio-temporal correlation and information entropy. By exploiting the spatio-temporal relationship of sensor data, the maximum correlation and weighted variation coefficient are introduced. The maximum correlation and weighted variation coefficient are used to analyze the state of the sensor nodes, and the confidence interval method is adopted to realize the adaptive update of the threshold. To improve the detection accuracy, the abnormal sensor node is confirmed by the linear least square estimation method using information entropy (IE-LLSE). The information entropy is adopted to analyze the data fluctuation to achieve the best prediction accuracy. Experimental results showed that the false positive rate and false negative rate of the proposed method were significantly reduced, and the accuracy was stable at more than 97%.