Research On Intrusion Detection Technology Based On Feature Optimization And BP Neural Network

The rapid progress of internet technology has promoted the fast development of all kinds of fields in society.Under this condition,network security problem has also followed.Network intrusion detection as one of the most important technologies which maintains the security of network system has been widely studied.There are still some problems in network intrusion detection such as mining incomplete features,too long modeling time,low detection rate and so on.In the paper,aiming at above problems,the data is processed after deeply analyzing and mining data of network intrusion behavior.And the BP neural network is used to predict and classify network intrusion behavior in order to improve the detection performance of network intrusion behavior.In the paper,the KDD CUP 99 network intrusion detection datasets which were collected by Lincoln Laboratory in 1998 is applied to analyze intrusion behavior.Firstly,the data types of dataset are classified and analyzed.The method of undersampling combined with oversampling is proposed,then the distribution of data types is balanced.And the integration method is adopted to extract important features.It is used to reduce the dimension of the data to obtain a simplified data subset for the subsequent construction of classifier model.Secondly,through the analysis of the BP neural network,the learning rate and iterations of BP neural network are optimized according to the weight adjustment process.Finally,the effectiveness of the optimized BP neural network is verified by experiments.In the paper,the NSL-KDD dataset is used for experimental comparison in order to prove the effectiveness of the proposed intrusion detection method.The method is based on feature optimization and BP neural network.The experimental results show that the proposed method has good performance in precision,recall,F-measure and other evaluation indicators.Compared with the unoptimized BP neural network algorithm,the modeling time of the presented model reduces nearly 8 multiples.The overall classification accuracy improves by 2%.It is particularly mentioned that the precision of U2 R and R2 L improves by 13.6% and 4.7% respectively.Compared with SVM and Naive Bayes classification methods,the overall detection accuracy of the suggested method is 2.9% higher than SVM and 11.8% higher than Naive Bayes.In addition,the detection rate of each data type is the best than SVM and Naive Bayes methods.The efficiency of the intrusion detection method is demonstrated based on feature optimization and BP neural network.