一种基于信息熵的传感器数据异常检测方法

An Anomaly Detection Method of Sensor Data Based on Information Entropy

传感器数据采集作为系统感知信息和获取数据的重要手段,其数据的真实性和可靠性至关重要,数据异常检测能提升数据的质量,挖掘出数据的潜在信息。基于分类、聚类等的检测方法依赖于数据的空间相关性,且复杂度很高,不适用于智能家居等小型物联网环境。基于数据距离的检测方法适用于此场景,但是存在误报率高的问题。针对这些问题,本文将传感器滑动窗口内的数据值作为离散随机变量,定义了数据流的信息熵,在此基础上提出了一种通过计算滑动窗口内信息熵进而检测数据异常的方法。模拟实验表明,本文提出的方法能高效地检测异常,并且有更高的检测率及更低的误报率,符合预期结果。

As an important means to perceive information and obtain data, it is crucial to make sensor datas ac curate and reliable. Data anomaly detection can improve the quality of data and the mining of potential information. The detection method based on classification and clustering relies on the spatial correlation of data, and the com plexity is very high. It is not suitable for smart home and other micro IOT environments. Besides, the detection method based on data distance is suitable for this scenario, but it has a high false positive rate. In order to solve these problems, the data values in the sliding window of the sensor are used as discrete random variables, and then the in formation entropy of the data flow is defined. On this basis, a method of anomaly detection for data in sliding win dow based on information entropy is proposed. Simulation experiments show that the proposed method can detect anomalies efficiently and has higher true positive rate and lower false positive rate, which is in line with the ex pected results.