Analysis And Research On Node Data Anomaly In Wireless Sensor Network

With the increasing application of Wireless Sensor Network(WSN),its data-centric and task-driven features are increasingly prominent.The main goal of the WSN deployment is to obtain data from the monitoring environment and extract valuable information from the data analysis,which is also the key to its successful application.However,data anomaly often occurs due to the network itself,hardware and software problems of node devices,and external harsh environments.Therefore,it is necessary to focus on and solve the two aspects of problems.First,design an effective data outlier detection mechanism to search for abnormal information from monitoring data.Second,a simple and efficient data recovery strategy is needed to deal with the error data those fail to objectively and truly reflect the monitoring environment itself.Based on the above considerations,the thesis conducts a deep research on the detection and recovery of abnormal data in wireless sensor network.The main research contents are as follows:1.Aiming at the shortcomings of ignoring spatial-temporal collaboration of the existing WSN node data anomaly detection methods,an abnormal data detection algorithm DADST based on valid close proximity and spatial-temporal cooperation is proposed.Firstly,the detection algorithm base on neighbors is improved,and a reliable neighbor screening model is established to filter the collaborated neighbor nodes with the reliability.Then the node data stability model is established to implement the data anomaly detection of the WSN nodes by using the spatial-temporal correlation of the monitoring data.Finally,simulation is designed to verify the validity of the algorithm by using the IBRL data set,which is published by the Intel Berkeley Laboratory.Experimental results show that the DADST algorithm can achieve high detection accuracy while maintaining a low false alarm rate.2.Aiming at the ignoring of node spatial correlation and abnormality duration in the existing WSN node anomaly data recovery methods,an improved anomaly data filling algorithm based on node similarity is proposed.Firstly,the traditional node data similarity measurement method is improved by introducing spatial distance factor,and a new node data similarity index is designed.Then,valid close proximity screening model is used to select the reliable referenced nodes those meet the similarity requirements for the data recovery.Then,according to the data recovery application scenario and the difference in the duration of the abnormalities,different numerical estimation methods are designed to recover the abnormal data.Finally,the effectiveness of the algorithm is verified from the data recovery residual size and distribution,the difference in the recovery effect between single measure and collaborative means by simulation,which is designed based on IBRL dataset.