Anomaly Detection Approach for Urban Sensing Based on Credibility and Time-Series Analysis Optimization Model

Urban sensor networks often consist of a large number of low-cost sensor nodes. Due to the constrained resource devices and hazardous deployment, urban sensing is vulnerable to interference and destruction of external factors or the impact of external environmental emergencies. Abnormal data, outliers, or anomalies have affected the utility in various domains seriously. Timely and accurate detection of unexpected events, monitoring of network performance, and anomaly detection of data flow are of great significance to improve the decision-making ability of the system. In this paper, we propose an anomaly detection method for urban sensing based on sequential data and credibility. First, based on Bayesian methods, a reputation model is established for the selection of credible sample points. Second, aiming at the problem that the threshold range is difficult to determine in the traditional method, the pivot quantity is defined by using the median of the credible sample, and the confidence interval can be estimated to quantify the deviation degree of the sensor data. Finally, an anomaly data identification and source verification approach is proposed to distinguish errors and events accurately. The evaluation results on both the detection rate and the false positive rate demonstrate a better performance of our approach than the other existing methods.