Research On Deep Learning Technology For Network Security Detection

Behind the rapid development of the Internet industry,Internet security has become a hidden danger.At present,the network security incidents occur frequently,the network attacks are diversified,and the network attacks tend to be "globalized" and "normalized".The issue of network security has become a common challenge facing the international community.How to understand the network security situation as early as possible,understand the development of network attack technology,and finally achieve effective network security detection has become the focus of public attention.Intrusion detection is an important and basic technology in the field of network security.It realizes network and computer system real-time protection in a manner of monitoring intrusion.How to strengthen the intrusion detection is our goal,but with the development of network attack technology,the traditional intrusion detection system has been unable to detect complex attacks accurately.In recent years,the outstanding performance of deep learning in classification and behavioral prediction based on massive data has led people to study how to use deep learning technology.Therefore,this paper attempts to apply deep learning to intrusion detection,learn and predict classification with the network attack behaviors.In this paper,according to the NSL-KDD dataset,we use the traditional classification methods and several different deep learning algorithms to conduct a study classification,and analyze the correlation between the data sets,the characteristics of the algorithms and the experimental classification effects deeply,after that find out a relatively good deep learning algorithm.Then,a normalized coding algorithm NSAE is proposed for the comparison experiment of NSL-KDD data set,which shows that the algorithm can effectively improve the detection accuracy and reduce the false alarm rate.The work of this paper is mainly reflected in the following aspects:1)Research on network attack behavior detection based on deep learning structure.Based on the Tensor Flow platform,the learning models of Softmax,CNN and SAE are designed.The model is implemented in Python language and the comparative experiments are performed on NSL-KDD datasets.The experimental results show that the network model under deep learning structure has better feature learning ability,and SAE has better learning effect than CNN algorithm.2)Based on the analysis of deep learning algorithms and the characteristics of intrusion detection data set,a normalized coding algorithm is proposed.The algorithm uses three hidden layers' codes and a normalized layer's learning network to make supervised learning with the intrusion detection data.The three hidden layers implement the extraction and encoding of the input information,the normalized layer implement the normalization of the result,and then conduct the regression learning according to the comparison of the target results.The experimental results show that the proposed algorithm can effectively improve the detection results,indicating that the work in this paper not only has theoretical significance but also has some practical significance for network security detection.