WSNs Intrusion Detection Method Based On LEACH Protocol And LightGBM Algorithm

Nowadays,with the rapid development of information technology,wireless sensor networks have been deeply investigated and widely used.Many security problems need to be solved in the application of wireless sensor networks.Because the intrusion detection system for wireless sensor networks is quite different from the traditional networks,the characteristics of sensor nodes should be fully considered to adopt the appropriate scheme,so that the network energy consumption is lower and the network life cycle is longer.The main work that has been done in this study includes:(1)With the help of WSN-DS data set,intrusion behaviors were analyzed.Firstly,the used experimental data were classified and analyzed to balance the data type distribution,and the RFE-SHAP feature selection method,which combined SHAP feature importance calculation method and recursive feature elimination method,was used to reduce the dimension of the data,thus effectively reducing the construction time of the classifier model.During the optimization of parameters in the model,various parameters of LightGBM algorithm and Optuna parameter optimization algorithm were investigated and analyzed,and important parameters of LightGBM algorithm were optimized.Finally,the improved parameter optimization algorithm was combined with the multi-classification model to establish the parameter optimization multi-classification model of intrusion detection in wireless sensor networks.(2)A routing protocol based on LEACH was proposed,which laid a foundation for meeting the energy requirements of intrusion detection.SOM algorithm was introduced to optimize the clustering of sensors in wireless networks.In addition,the strategy of node data transmission in the sensor network was changed.In order to verify the effectiveness of the scheme,Matlab software was used to simulate the information exchange mode of sensor nodes,and the improved scheme was compared with the traditional wireless sensor networks routing protocol in terms of cluster distribution,the number of surviving remaining nodes and the total remaining network resources.The survival time of sensor nodes and the overall running time of the network were extended by 34%.The improved routing protocol proposed in this study can fully reduce the energy consumption speed of wireless sensor networks,increase the data throughput and extend the network life to a certain extent.(3)The WSN-DS data set was used for experimental to verify the reliability of the intrusion detection scheme of wireless sensor networks proposed in this study.The experimental results showed that the proposed method had good performance on different indicators.The modeling time of the proposed algorithm was 46% shorter than the traditional LightGBM algorithm.Compared with GBDT,SVM,Random Forest,XGBoost and CNN,the overall detection accuracy of the proposed method was 1.28%,10.66%,0.97%,0.57% and 0.28% higher than that of GBDT,SVM,Random Forest,XGBoost and CNN,respectively.The detection rate and modeling time for different categories of data were much better than the above classification methods,which verified the effectiveness of the wireless sensor networks proposed in this study in intrusion detection.