| Network anomaly detection is a key technology to improve network security,has received extensive attention from researchers,its network protection adopts active defense strategy,which is of great research value for improving the detection performance of network anomaly detection system.At the same time,because network security affects every field of people’s daily life,the research on network anomaly detection also has very important application value.Therefore,building a network anomaly detection model with high detection accuracy and fast detection speed has become the main problem to further consolidate the network security barrier.First of all,the general process of the network anomaly detection system is outlined,and the composition and data characteristics of the network data are analyzed,focusing on two deep learning algorithms,the convolutional neural network and the long-shortterm memory network,and building a network anomaly detection model based on this.In view of the large gap in the number of samples among different categories in the network data set,research on data enhancement algorithms is carried out,and in view of the problem of slow detection speed caused by excessive data volume and complex sample features,research on feature selection methods is carried out,and several commonly used evaluation metrics for evaluating model checking performance are given.Secondly,from the perspective of improving the detection accuracy of network anomaly detection model,proposing a network anomaly detection method based on data enhancement.Aiming at the problem that the number of minority samples in the training set is too small and the model cannot effectively learn the features of minority samples,the number of minority samples is increased by data enhancement method,so that the model can learn the characteristics of different types of samples in the training process,so that the model can obtain better detection performance on the test set,and the application scenario of the method is given through the comparison experiment.Further,for solving the problem of slow detection speed caused by the increasing number of samples and complex feature composition,proposing a network anomaly detection method based on feature selection.Realizing the dimensionality reduction of the original data set by the feature selection method.The reduction of data feature dimensions effectively improves the detection speed of the model.The proposed method can effectively reduce the model training time and greatly reduce the test time.Compared with the feature selection methods in other literatures,it can obtain a higher detection accuracy,indicating the superiority of the method in the process of network anomaly detection.Finally,comparative experiments are carried out in the binary classification and multi-classification tasks,and the experimental results of multiple experiments are analyzed,the advantages and existing problems of the methods proposed in this paper are given,and the future work direction is clarified.In the comparative experiment with the existing methods,it is found that the proposed method has high detection performance in the process of network anomaly detection,which can meet the requirements of network anomaly detection system for detection accuracy and detection speed,and further enrich the network anomaly detection technology. |
