Research On Data Anomaly Detection Algorithms Based On Spatiotemporal Correlation In Wireless Sensor Network

Wireless Sensor Networks(WSNs)is widely used in multiple fields for data collection and monitoring due to their advantages of low cost,easy deployment,low energy consumption,and small size.However,resource constraints and instability of sensor nodes can lead to the collection of a large number of data with deviation from actual characteristics,which will affect the lifetime and data reliability of wireless sensor networks.In addition,sensor nodes will generate abnormal readings based on environmental changes,which required timely identification and early warning.Research on anomaly detection technology can effectively improve the stability of the network and the authenticity of data,which is the core content of wireless sensor networks.This article is based on the temporal correlation of data collection and the spatial correlation of deployed nodes in the network.It focuses on how to effectively identify anomalies while reducing the time cost of detection,as well as how to accurately identify event anomalies and reduce communication costs.The main work is as follows:(1)In order to solve data timeliness requirements and the problem that nodes in the network are limited in resources and cannot perform complex computation of big data,an algorithm named multi-attribute data anomaly detection method based on edge computing(MDADE)is proposed.The sensor node performs feature transformation on attributes based on Pearson correlation coefficients,reducing data dimensions,and improving computational efficiency.It uses sensor node anomaly detection algorithms to eliminate noise anomalies caused by software faults and reduces redundant data transmission.The sink node introduces sliding window technology to process the data flow according to time series,performs clustering anomaly detection algorithm on the data uploaded by multiple sensor nodes within the same time period,eliminates anomalies caused by hardware failures,and identifies contextual anomalies in the data flow through data change trends.The simulation results show that the MDADE algorithm can identify abnormal data effectively and reduce the time overhead significantly.(2)Aiming at the problem that sensor nodes are susceptible to deployment environment,resulting in erroneous readings that affect event anomaly detection,a neighborhood relationship-based data anomaly detection algorithm(NR-DAD)is proposed.NR-DAD performs internal anomaly detection on sensor data at nodes to identify abnormal data.Dividing neighborhoods based on perceptual range coverage ensures spatial correlation between nodes and neighborhoods.Determine the type of anomaly by predicting the node status and analyzing the collaborative perception of neighboring nodes.Reducing longdistance transmission times and data volume in the network through close range transmission and packet sending.Experimental results show that the NR-DAD algorithm can distinguish between event and error effectively,while having lower network energy consumption and high network lifetime.