Research And Application Of Network Intrusion Detection Method Based On Feature Selection And Ensemble Learning

In recent years,with the development of the Internet,the network has penetrated into people’s daily life,entertainment,scientific research and other fields.However,with the development of the internet,the importance of network security has become increasingly prominent,accompanied by various frequent network security events.Machine learning based intrusion detection,as one of the core technologies of network security,can effectively identify various attacks in the network and provide timely warning feedback,providing reliable technical support for network security defense.Considering the increasing scale and complexity of current network data,new attacks emerge one after another and become more hidden and difficult to detect,this paper studies three aspects:feature reduction and optimization of network data,integrated strong classifier,application of the proposed method and combining misuse detection and anomaly detection to design a hybrid cascade detection model framework.The main content includes the following aspects:To reduce the dimensionality of network data feature attributes,a dual feature selection algorithm based on correlation improvement and Helly hypergraph is proposed.Firstly,the improved correlation algorithm ICFS is used to reduce the dimensionality of data features,removing a large number of irrelevant and redundant features to obtain a preliminary feature subset.Then,the Helly hypergraph is used to optimize the preliminary feature subset and minimize the feature subset to obtain the optimal feature subset.In addition,design intrusion detection multi classification methods based on OVR strategy.Theoretical research and experiments have shown that the proposed method can effectively reduce the dimensionality of data features and obtain the optimal simplified feature subset,which can improve the accuracy and efficiency of detection.A network intrusion detection method based on ISVM integrated multiple classifiers is proposed to address the new changes in complex network detection environments,stronger attack concealment,and higher randomness.Three kinds of classifiers with different detection principles are selected as the base classifier.The improved PSO algorithm is used to optimize the parameters of SVM and as a meta-classifier.After multi-feature subset division and training,the results are integrated with ISVM.Through experiments and comparison with other methods,the proposed method can improve the accuracy of intrusion detection and to some extent improve the detection of new attacks.In order to make the network intrusion detection system have a certain self-learning ability,take into account the detection accuracy and the detection rate of new attacks,a hybrid cascade network intrusion detection model framework based on machine learning is proposed.This model framework combines misuse detection and anomaly detection through a hybrid cascade approach,which can leverage the advantages of each detection technology.After experimental verification,the model framework has selflearning ability and can continuously improve the accuracy and generalization ability of the detection system.