Research On Intrusion Detection Mechanism Based On Modular Neural Network

In recent years,with the rapid development of Internet in China,network technology continues to improve our lives.Due to its own structural characteristics,the traditional intrusion detection model will show the problems of weak expansion ability,poor test stability and slow detection speed when the amount of network data increases.Modular neural network is a network structure composed of modular concept.The model has the characteristics of simple network structure,strong expansibility and high computational efficiency.Aiming at the problem of intrusion detection,this paper improves the network structure of modular neural network,and designs a modular neural network model for intrusion detection.(1)In the task decomposition part,the Meanshift-Kmeans algorithm is proposed.According to the characteristics of intrusion detection data,the algorithm is used to decompose the task and divide the data into different clusters.(2)In the part of module selection,similarity selection is proposed.The result of decomposition is used to judge the similarity,and the dynamic sub module selection is carried out.It lays the foundation for sub tasks to be assigned to sub modules.(3)In the sub module,the neural network is replaced by a classifier,and the optimal detection model is selected to deal with this kind of attack through data experiments in different classifiers.In the part of sub network integration,according to the results of subtask processing,a sub network integration method similar to neural network is proposed.According to the results of data processing,the integration of sub modules is realized.In this paper,the intrusion detection model based on modular neural network is compared with other machine learning models,Through the comparative experiments on two different intrusion detection datasets,the good performance of the modular neural network existing in this paper is demonstrated,and the efficient detection ability of the model is confirmed.