Research On Network Intrusion Detection Technology Based On Deep Learning

With the emergence of new Internet technologies such as file sharing,mobile payment and instant messaging,the network security situation is becoming more and more complex,and at the same time,network attackers become more covert,seriously threatening the network security environment,making the problem of computer network security has become a major concern.There are many network security protection measures,such as firewall technology,intrusion detection,antivirus software,etc.,among which intrusion detection technology is one of the key technologies for information security guarantee.However,due to too much human involvement in the traditional intrusion detection system,the detection accuracy is low and it is not intelligent enough.In the case of actual detection,due to the unbalance of intrusion data,unbalanced data will occur in the detection,and high-dimensional and unbalanced data bring great challenges to the intrusion detection system.In view of the above problems,this paper first introduces the basic knowledge,technical classification and existing problems of intrusion detection system.Then,the basic knowledge of deep learning and its classical model are analyzed.Finally,the main methods of unbalanced data processing are briefly summarized.After a series of intrusion detection investigations under high-dimensional and unbalanced data,it is decided to use Adversarial autoencoder(AAE)integrated learning based on Borderline-strikes to detect the problem of large and unbalanced data in intrusion detection system.The main part works as follows:(1)Aiming at the problem of sparse data distribution,redundancy and irrelevant features caused by high-dimensional data,In this paper,the Adversarial autoencoder algorithm with excellent dimensionality reduction function is used to perform dimensionality reduction,the effectiveness of the proposed scheme is verified by experiments.But this step only solves the effective mapping from high dimensional data to low dimensional data in intrusion detection,The Adversarial autoencoder algorithm is not ideal for detecting unbalanced data.(2)Aiming at the problem that data traffic caused by unbalanced data is difficult to detect.Although integrated learning has achieved good results in dealing with unbalanced data,it has some limitations in dealing with high-dimensional data.Therefore,in this paper,the combination of the Borderline-SMOTE oversamping and the EasyEnsemble integrated learning is adopted to deal with the unbalanced data and solve the problem of high false positives caused by the unbalanced data.(3)Aiming at the complex problem that the effective information brought by the high-dimensional and unbalanced data is difficult to obtain and the information of a few attacks is difficult to identify,This paper presents an intrusion detection technique based on Borderline-SMOTE oversampling and Adversarial autoencoder integrated learning.The comparative experiments of SMOTE-Borderline-AAE,EasyEnsemble-AAE and SMOTE-Borderline-EasyEnsemble-AAE verify the superior performance of the proposed scheme.The experimental results show that the network intrusion detection model based on Borderline-SMOTE AAE integrated learning proposed in this paper can identify the network attack behavior more accurately,in the actual application environment,it has more efficient use value.