Research On Anomaly Detection Method Of Wireless Sensor Network Based On KPCA And SVDD

With the rapid development of wireless communication technology and sensor technology,Wireless Sensor Network(WSN)with sensing function,computing function and communication function has started to be widely used.Due to the limitations of the sensor nodes and the uncertainty of the external environment,the data collected by WSN are prone to noise or anomalies,which can have a huge impact on data analysis.Data in real environments usually have high dimensional characteristics,and suitable outlier detection algorithms need to be proposed for different data characteristics.In this paper,we study anomaly detection methods for high-dimensional data in wireless sensor network,and the main work is as follows:(1)To address the limitations of Principal Component Analysis(PCA)in dealing with nonlinear high-dimensional data,a feature extraction algorithm based on Kernel Principal Component Analysis(KPCA)is proposed in this paper.By introducing the KPCA algorithm with kernel function to replace the traditional calculation method in high-dimensional space,the feature extraction effect is greatly improved compared with the PCA algorithm.Considering the randomness of parameter selection in the kernel function,we continue to use the particle swarm optimization algorithm to perform global optimization of the parameters in the kernel function and use the optimal parameters to build the outlier detection model.In order to effectively detect outliers in high-dimensional data and adapt to the dynamic changes of data,the established model is continuously updated to ensure its applicability and accuracy.The experimental results show that the proposed algorithm can retain more initial information and effectively realize the dimensionality reduction of high-dimensional data features,while improving the accuracy of anomaly detection.(2)To address the problems of online real-time detection and poor detection accuracy in data anomaly detection,this paper proposes a data anomaly detection algorithm based on the sliding window model and Support Vector Data Description(SVDD).The algorithm first adopts KPCA algorithm for high-dimensional data feature extraction to realize the dimensionality reduction of high-dimensional data,and then uses SVDD algorithm with linear kernel function under the sliding window model for anomaly detection.The experimental results show that the sliding window model can realize online real-time data detection and improve the accuracy of the anomaly detection algorithm,and the detection accuracy of the proposed linear kernel function SVDD algorithm reaches 97.8%,which is a significant improvement in accuracy compared with other algorithms,and the computing time of the algorithm is also significantly reduced.(3)In response to the problem of not being able to control the process of WSN anomaly detection and visualization of detection results,this paper designs and implements a wireless sensor network anomaly detection system.Firstly,the overall requirements of the wireless sensor network anomaly detection system are analyzed,and the overall platform of the system is designed through the requirements analysis.The system is divided into three parts: browser side for user and data interaction process,server side for business logic processing,and database for data storage.Each part is designed independently to separate the data from the business logic.The detection system enables the transmission and storage of node data,the loading and training of anomaly detection algorithm models,and the display and interaction of operation results,providing an effective way for users to perform anomaly detection in wireless sensor network.