Research On Intrusion Detection Algorithm Based On Multi Model Integration

With the advent of the era of big data,network intrusion events emerge in endlessly,and network security has received more and more attention.For example,traditional network security such as firewall often has problems such as low detection rate and rigid methods,which can not protect our network security in many ways.Compared with the traditional network security technology,intrusion detection has the huge advantage of active defense against unknown attacks in the security field.Active defense against unknown attacks has always been an important driving element in the development of intrusion detection.However,due to the limitations of the algorithm itself,the intrusion detection system is controversial,and the accuracy of a single algorithm model is not high,and the generalization performance is poor.In order to further explore the methods to improve the efficiency of intrusion detection,this paper studies the intrusion detection algorithm.Through the in-depth exploration of intrusion detection and machine learning knowledge,aiming at the problems of intrusion detection data and technology,this paper proposes an intrusion detection algorithm based on multi model integration.Firstly,the characteristics of intrusion detection data set are analyzed,and the KDDCUP99 data set is preprocessed to solve the problems of different dimensions,noise and discretization in the data set,and the data needed to process the original data into machine learning model is obtained,so that the data set can be input into the model without losing information.Aiming at the problem that the number of features in the dataset is too large,a variety of dimensionality reduction methods are experimented.PCA,LDA and tree based feature selection are used to reduce the dimension of the data set after data preprocessing.Combined with the model detection effect,the appropriate dimension reduction method is selected.Finally,the standard data set is obtained by using this dimension reduction method.Secondly,the appropriate machine learning model is selected for modeling.Six kinds of algorithm models are trained and predicted,including logistic regression algorithm,extreme random tree algorithm,Ada Boost model,multi-layer neural network algorithm,decision tree algorithm and random forest algorithm.The machine learning model with the best effect and completed training is reserved.According to the prediction results,the shortcomings of each model are analyzed to prepare for model integration.Finally,the key step of this paper is to select the appropriate multi-model integration strategy.In this paper,the reserved model set is introduced into the integration algorithm.Combining the advantages of each algorithm model and using the integration strategy of multiple models,an intrusion detection model based on multi model integration is designed.The experiment and prediction of intrusion detection model show that the multi model integration algorithm can effectively improve the detection rate of intrusion detection,and the integrated model has better generalization ability.Figure 48;Table 16;Reference 52...