Research On Intrusion Detection Technology Based On Convolutional Autoencoder And Long Short Term Memory Network

The current network environment is becoming more and more complex,a large amount of network traffic puts forward high requirements for intrusion detection systems,and attackers’ attack methods are becoming more and more subtle and diverse.Therefore,in response to these challenges,the intrusion detection system needs to have efficient data processing and analysis capabilities,powerful data mining capabilities,and the ability to identify different attack methods to adapt to the ever-changing network environment and attack methods.The deep learning intrusion detection model can automatically learn and extract advanced features in the data through a multi-layer neural network,which is more efficient than traditional machine learning methods.Moreover,the performance of the deep learning model is much higher than that of the machine learning model,and it has stronger generalization ability,so it can better adapt to different types of intrusion behaviors and has better versatility,the following improvements are made to the current intrusion detection model:(1)In view of the high-dimensional and redundant phenomena presented by the current network traffic data,an intrusion detection model based on convolutional autoencoder and long short term memory network is proposed.By learning the spatial and temporal characteristics of network traffic,Comprehensively extract the key feature representation of network traffic and improve the feature discrimination between various types of network traffic,compared with other popular models,the performance is even better.(2)In view of the data imbalance problem existing in the current intrusion detection training data set,it is easy to cause the overfitting of when training the model,and it is difficult to improve the classification accuracy of minority network traffic.Therefore,a method based on generative confrontation network is proposed.The method increases the proportion of minority class network traffic and reduces the risk of model overfitting.Compared with other generated intrusion detection models,the overall classification accuracy of the model is stronger.(3)The current mainstream intrusion detection model has a high ability to identify known types of network attacks in the intrusion detection data set,but the ability to identify unknown types of network attacks is insufficient.In this paper,the discriminant conditional variational autoencoder is used to improve the convolutional autoencoder,to reconstruct the normal network traffic,and use the density peak clustering algorithm to find out the distribution of normal network traffic,which can help the intrusion detection system to identify unknown network attacks,and has stronger intrusion detection performance in the face of complex network environments.Experiments show that the intrusion detection method proposed in this paper performs well in the classification experiments of the NSL-KDD data set,and in the classification of minority attacks and unknown types of network attacks,the detection rate and accuracy are significantly improved,and the overall performance of the model has been improved.