Research On Intrusion Detection Based On Deep Belief Network And SVM

In the network security information system,the intrusion detection technology protects the user's information security from the illegal elements.In the past,intrusion detection system mainly learned from bitbase or by database.The main point is how to select the intrusion features exactly.Only a deeper understanding of the features can make the IDS have excellent performance and high efficiency.Modern intrusion detection is faced with a great challenge that it is difficult to detect variety attacks,and the detection rate and accuracy rate are not high enough.In order to meet the above challenges of intrusion detection,in this paper,deep learning and traditional IDS are combined to solve the above test.This paper designs and implements an intrusion detection system model which combines DBN and SVM.The data set adopts the nsl-kdd data set commonly used in the field of intrusion detection.Redundant cleaning,data type conversion,normalization and other processing operations are carried out for the data set.The standard data set with low redundancy and low dimension can be obtained by using PMF coding.Due to the excellent performance of deep learning in feature learning,using the DBN to learn data sets can essentially extract the low-dimensional features of network attacks,improve the adaptability of the model to new attacks,and essentially distinguish whether network behaviors are invasive.SVM classifier has a strong ability of classification,especially for low-dimensional features.Therefore,SVM is selected to classify the features after dimension reduction of attack behavior,which can better improve the detection rate of the system.Parameters of the whole dbn-svm model are optimized,such as batch size,number of nodes,number of network layers and so on,in order to obtain the optimal detection performance.After the optimization,the test set data is tested by the model and compared with the detection ability of the traditional model.The experimental data shows that the intrusion detection system based on dbn-svm model can effectively detect and classify the attack behaviors in the data set,which is significantly improved compared with the traditional model.It is proved that the model based on dbn-svm is helpful to improve the detection ability of IDS.