Design And Implementation Of Network Anomaly Access Detection System Based On GAN

In the new era,the Internet has become the key to the national development strategy.The Internet is more and more closely related to people's life,so the network attack detection work is facing great challenges.The accuracy and timeliness of network anomaly detection determine the availability and reliability of the current network.However,in the face of high-speed and large-scale Internet environment,the network data dimension is increasing,and the network attack means are more and more hidden.Many traditional network anomaly access detection technologies can not detect network anomalies efficiently and accurately,and it is difficult to meet the requirements of network security.Generative adversarial network(GAN)is a model framework based on deep learning,which adopts adversarial strategy mechanism.It can make up for the deficiency that traditional methods need a lot of abnormal data in the training process.Therefore,this dissertation proposes an effective method to detect abnormal network access based on GAN.First of all,aiming at the problems of feature redundancy and missing data values in network abnormal access data,this dissertation analyzes the data features of other samples without missing to preprocess the data.The main methods are symbolic feature digitization,missing values filling,data values standardization and data values normalization,and turn it into data that can be used in subsequent experiments.Secondly,the network anomaly access detection framework based on generative adversarial networks is proposed.The framework generates "false" data by generators in generative adversarial network,which solves the problem of less abnormal data in traditional anomaly detection.In order to solve the problem that the model is difficult to train in the original GAN,two kinds of derivative models of generative adversarial network are designed by this dissertation: one is to replace the discriminator and generator in the original GAN with the long short-term memory network;the other is to add the encoder into the generative adversarial network.The experimental results show that the performance and stability of the network anomaly access detection algorithm based on generative adversarial network are significantly improved compared with the traditional algorithm.Compared with the traditional method,the detection method based on GAN has 1% improvement,the model based on long short-term memory network has 1% improvement,and the model based on encoder has 15% improvement.Finally,the function modules of the system are analyzed and designed,and the network abnormal access detection system based on generative adversarial network is developed,and the results of the system are further visualized.The system can update all kinds of network anomaly types and all detected anomaly data in real time,which is helpful for theoretical demonstration and experimental analysis in the next step.