Research On Network Intrusion Detection Technology Based On Convolutional Neural Network

With the development of new Internet technology and the expansion of the network scale,the convenience that the Internet provides to people's lives is ubiquitous and has become an indispensable part of people's production and life.However,at the same time,various cyber security problems have become more and more serious,and even premeditated cyber attacks and cyber crimes have brought huge challenges to our cyberspace security.In the increasingly severe network security situation,network intrusion detection,as an active defense technology,can detect the network and find abnormal behaviors in the network,which plays a very important role in realizing the early warning and dynamic defense of abnormal behaviors in the network.At present,the common detection technology is the intrusion detection model obtained by using the traditional machine learning method to train the intrusion samples.However,there are some deficiencies at present,such as low detection rate,only dealing with manually labeled standardized data,and little research on the original network traffic data processing.The main work of this article is as follows:1.In view of the low accuracy of intrusion detection models based on traditional machine learning,this paper designs a network intrusion detection model based on convolutional neural network algorithm.The model uses a deep separable convolution layer to improve the traditional convolution layer.By discarding the original convolution operation,two separable convolutions are used for feature extraction to reduce model training parameters.Long Short-Term Memory neural network retains the sequential structure of features,which can fully extract the effective features of the invading samples,so as to accurately classify the invading samples.And the model is applied to the processing of KDDCUP99 data set in the field of intrusion detection.The experimental results show that the accuracy of the model in the KDDCUP99 data detection can reach 98.7%.2.Aiming at the processing of unlabeled original network traffic data,this paper designs a multi-model fusion network intrusion detection method.The traditional machine learning method is used to better classify the standardized data.On the one hand,the normalized raw network traffic is first divided into normal and abnormal traffic by the SVM method.The improved convolutional neural network model is used to detect only abnormal traffic.On the other hand,the normalized raw network traffic is divided into normal and various abnormal traffic by the KNN method.And the two processing results are merged to improve the model detection efficiency and detection accuracy.In this paper,the original network traffic data set CIC-IDS-2017(Pcap)is converted into standardized CSV format data as the input of the model.The experimental results show that the intrusion detection model based on convolutional neural network has higher accuracy and more practicability than the traditional machine learning technology in detecting the original network traffic data.