Network Anomaly Detection System Based On Deep Learning

Nowadays,Internet technology has penetrated into all aspects of people’s lives.Artificial intelligence products such as Tmall Genie,Siri and sweeping robots have become common scientific and technological products.The Internet technology has greatly improved people’s quality of life.But at the same time,because of the popularity of the Internet,hackers,malware attacks and other intrusion behaviors emerge endlessly,even in the national level,it has even posed a huge threat to our country’s security at the national level.Therefore,it has become an important research part in the field of network security that how to detect abnormal behaviors in network efficiently.The correct detection of abnormal network data can effectively solve the problem of network security.Due to the characteristics of large amount of network data and high dimension,despite the network anomaly detection technology based on machine learning is becoming more and more mature,the network anomaly detection technology based on machine learning still has the disadvantages of low detection rate and high false alarm rate.And deep learning is superior to machine learning in processing large-scale data.In view of this situation,it is necessary to vigorously research and develop a network anomaly detection system based on deep learning.This paper mainly focuses on the characteristics of network traffic data,designs a network anomaly detection model by applying deep learning technology to the field of network anomaly detection,and based on this model,a based on deep learning is implemented.The network anomaly detection system proposed in this paper uses the KDD99 data set as the experimental data set.In the part of feature selection,the one-off method is compared with the method combined with random forest and RFE to select the better one as the final feature selection method.In the part of model selection,after comparing BP neural network model,convolution neural network model and Bi-LSTM network model,the three models are compared with the model based on SVM,and finally the best model is selected as the core model of the system.The system implemented in this paper has four modules,which are network data collection module,network data processing module,prediction and result display module and user management module.The network data collection module uploads network packets and extracts effective features from them.The network data processing module is to keep the data format consistent with the model training data.The prediction and result display module predicts whether there is abnormal behavior in the network through the designed network anomaly detection model,and then outputs the category.User management module is designed for the administrator to manage ordinary users.Finally,the experimental results show that the effect of network anomaly detection model based on Bi-LSTM is better than the BP neural network model and the model based on SVM,so it can be proved that deep learning has good performance in network anomaly detection.