Font Size: a A A

Design And Implementation Of Intrusion Detection Method Based On Deep Belief Network

Posted on:2020-05-12Degree:MasterType:Thesis
Country:ChinaCandidate:X ZhangFull Text:PDF
GTID:2438330623464150Subject:Computer technology
Abstract/Summary:PDF Full Text Request
The development of network technology not only brings convenience to people's lives,but also makes the problem of information security increasingly prominent.Intrusion detection technology is an important part of the information security discipline,and it is an active and effective dynamic network defense technology.The traditional intrusion detection technology has some problems,such as poor detection performance and low adaptive ability.And because of the rapid development of the network,the volume and dimension of network data are increasing day by day,and there are a lot of unlabeled data in the network.Therefore,the traditional intrusion detection technology has been difficult to adapt to the current complex network environment.Deep Belief Network(DBN)is a cutting-edge technology in the field of deep learning.It has certain advantages in feature learning,and its model training adopts unsupervised learning method with layer-by-layer initialization,which can effectively utilize a large number of unlabeled data in the network.And it has achieved remarkable achievements in speech recognition,natural language processing and other fields.Based on this,this thesis discusses the application of deep belief network in the field of intrusion detection in order to solve the current network security problems.The following is a brief description of the work done in this thesis:1.Aiming at the limitation of BP network on the top layer of DBN detection model to the model detection accuracy,this thesis constructs an intrusion detection model that combines the Deep Belief Network and Extreme Learning Machine: DBN-ELM detection model.This model combines the advantages of deep belief network in feature learning with the advantages of extreme learning machine in single hidden layer feedforward neural network learning,which effectively solves the deficiencies of DBN detection model.Based on the NSL-KDD data set,this thesis constructs a comparative experiment between the DBN-ELM detection model and the DBN detection model.The experimental results show that compared with the DBN detection model,the DBN-ELM detection model has been improved in both false alarm rate and accuracy.2.Aiming at the low accuracy of the DBN-ELM detection model for unknown intrusion detection,based on the DBN-ELM detection model,this thesis proposes an intrusion detection model that combines deep belief network(D),extreme learning machine(E)and feature element selection and combination(R): DRE detection model.Before ELM module detects and identifies data features,this model can select and combine the elements of the original data features extracted by DBN module to obtain the optimal combination of feature elements of the original data.At last,based on the NSL-KDD data set,this thesis conducted an experimental analysis of this model,and constructed a comparative experiment between it and the DBNELM detection model.The experimental results show that the DRE detection model improves the detection accuracy of unknown intrusion,and it has complementary advantages with the DBN-ELM detection model in classification and recognition of network behavior.Therefore,in practical application,the two detection models can be combined to improve the adaptability of intrusion detection system to the current complex and high-dimensional network environment.
Keywords/Search Tags:Intrusion detection, Feature learning, Deep belief network, Extreme learning machine, Selection and Combination of Feature elements
PDF Full Text Request
Related items