Research On Network Traffic Anomaly Detection Method Based On Deep Learning

With the development of computer networks,network technology is flourishing and growing,more and more network applications are popularized in people's daily lives,network security problems are also increasing,and attackers on the network are also constantly improving Attack technology makes it more difficult to detect abnormal network behavior.Traditional abnormal traffic detection methods can no longer meet the security needs of the Internet.In recent years,the network abnormal traffic detection technology based on deep learning has satisfactorily met the security needs of the Internet,not only for dynamic prevention,but also the need for human intervention from the input of the original data to the output of the final result.Compared with traditional machine learning algorithms,various algorithms for deep learning have more obvious advantages in massive data processing.Therefore,in view of the shortcomings of traditional network traffic anomaly detection,this paper proposes a network traffic anomaly detection method based on deep learning.First of all,we cut,align,and complement each request traffic in the data set to generate a series of matrix data A as input,and then improve the pooling layer of the convolutional neural network to build dynamic adaptive pooling(DAPA),So that it can dynamically adjust the pooling process according to different feature maps.Use the data set for model training,and add a Dropout layer to the network structure to solve the problem of overfitting in the process of traffic feature extraction.Use softmax function to perform binary classification on test set data.First of all,a series of comparative experiments were carried out on the algorithm model of this part.The data set uses the publicly generated data set HTTP CSIC2010,which contains tens of thousands of request data.The experimental results show that the accuracy of using dynamic adaptive pooling is significantly improved and the loss value is reduced,and the problem of overfitting is also solved.The algorithm is an efficient algorithm that can be used for traffic anomaly detection,with certain accuracy and feasibility.Based on the feasibility and accuracy of the above dynamic adaptive pooling convolutional neural network model,a network abnormality monitoring system based on DAPA-CNN is designed and implemented.In Chapter 4,the entire system is described in detail,including the overall structure of the system and the specific process flow of the system.At the same time,various functions of the system were verified through experiments.The experimental results show that this system is feasible and highly accurate in detecting and monitoring abnormal network behaviors.