Research On Intrusion Detection Model Based On Dimensional Reduction And Improved CS-WNN

Due to the complex and varied types of cyberspace attacks,intrusion detection technology,as an important network security technology,is constantly experiencing new challenges.In order to meet the double requirements of high precision and fast intrusion detection speed,this paper focuses on in-depth research on intrusion detection technology,intrusion detection benchmark dataset,Wavelet Neural Network(WNN),Cuckoo Search(CS)and data dimension reduction methods.Then an intrusion detection model based on data dimension reduction and improved CS-WNN is proposed.The main work of this article is as follows:The commonly used intrusion detection benchmark data sets are old and cannot effectively reflect the real network traffic today.This paper uses the newly proposed NUSW-NB15 data set as the intrusion detection data set,which makes the model more practical.The NUSW-NB15 data set was further studied and compared with the commonly used KDDCUP99 data set.The advantages of the NUSW-NB15 data set were analyzed in detail and various methods were adopted to make it more suitable for the research of intrusion detection model.For the problem that the gradient descent algorithm in Wavelet Neural Network is easily affected by the initial parameters and the traditional Cuckoo Search has poor local search capability,and the search is likely to fluctuate later in the search,this paper proposes an improved Cuckoo Search and uses it to optimize the initial parameters of Wavelet Neural Network.And an improved CS-WNN intrusion detection model is proposed.The simulation and comparison experiments on the NUSW-NB15 dataset verify that the proposed algorithm improves the optimization performance of the model.In order to solve the problem of high dimensionality and large amount of NUSW-NB15 datasets,Laplacian Eigenmaps and feature selection methods are respectively adopted to reduce the dimension of the dataset,and the reduced-dimension dataset is applied to the improved CS-WNN intrusion detection model.The dimensionality reduction effect of different datadimension reduction methods was compared by simulation experiments,and the dimension reduction results were analyzed,and the best way is chosen.The simulation experiment results show that the data dimension reduction and the improved CS-WNN intrusion detection model proposed in this paper have good accuracy,robustness and generalization ability.