Study On Intrusion Detection Method Based On Rough Set Attribute Reduction And Bayesian Classification

Protecting the computer network information system and building a reliable information security protection system has become an important issue in the information society.In recent years,the application of machine learning methods to intrusion detection technology has attracted more and more attentions.The intrusion detection method using machine learning improves the detection efficiency of the system and at the same time reduces the false alarm rate and the false alarm rate of the intrusion system.Among machine learning algorithms,the Bayesian classification method has its own advantages when processing sample data sets.The principle is simple,the training process is fast,and the results obtained are intuitive and easy to interpret and understand.However,when processing high-dimensional data samples,too many attributes actually affect the training process of the classifier.In addition,the classification training time is too long and the performance is poor.In this paper,the combination of Bayesian algorithm and rough set attribute reduction algorithm is applied to intrusion detection method.While improving the reduction speed,it can improve the classification efficiency under the premise of improving accuracy rate.The main work is as follows.This paper investigates to present an improved attribute reduction algorithm.The basic idea of the algorithm is that attribute importance is defined by attribute dependency.As long as the calculation of attribute dependency can be accelerated,the speed of attribute reduction can be improved.An improved and fast attribute dependency calculation algorithm is given.On this basis,a fast attribute reduction algorithm is designed and implemented;a parallel algorithm based on improved dependency calculation and attribute importance calculation is designed and implemented,and then a fast Bayesian classification intrusion detection algorithm based on the importance of rough set attributes.The algorithm can reduce the dimension of high-dimensional data,remove redundant features,improve the accuracy of intrusion detection,and improve the algorithm speed and detection efficiency.Based on the KDDCUP99 data set with about 4.9 million records,three groups of experiments were compared.Experiments show that the Bayesian classification method is superior to the SVM classifier and the KNN classifier in scenarios that require high detection response speed.Ten-fold,five-fold,and three-fold cross-validation tests show that the improved Bayesian classifier in this paper is superior to the traditional naive Bayes classifier from accuracy to running time.Experiments show that the improved intrusion detection algorithm based on attribute importance is higher than the other three intrusion detection algorithms based on attribute reduction.