Research On Network Intrusion Detection Based On Deep Learning Model

In the Internet age,computers are closely related to people's lives,which have greatly improved people's lives,but there are also many dangers.The scale of network data is getting larger and larger,and many threat behaviors in the network have brought huge hidden dangers to network security.The network intrusion detection system is an active defense method to ensure network security.real-time response to abnormal data.Although traditional machine learning algorithms also have obvious effects in intrusion detection,because they rely too much on feature engineering and feature selection,as network data increases day by day,it is impossible to keep intrusion detection real-time and accurate.Over the past few years,deep learning has achieved very good results in fields such as computer vision and natural language processing.In order to meet the requirements of network intrusion detection,this paper studies a network intrusion detection method based on deep learning model.The main work is as follows:In order to improve the autonomous defense capability and correct detection rate of the network intrusion detection system,this paper proposes an intrusion detection model that combines the convolutional neural network and the Inception network structure,and sets the attention mechanism in the model.Drop Block layer.This model uses the convolutional neural network layer to fully extract data features,calculates the weight of each feature through the attention mechanism to distinguish the importance of the feature,and uses the Drop Block layer to improve the generalization ability of the model,which improves the accuracy of intrusion detection.At the same time,the complexity of the model is reduced.Experiments on KDD99 dataset show that this model has higher accuracy and stronger generalization ability.In order to improve the detection efficiency and accuracy,and fully learn the characteristics of a small amount of data,this paper first analyzes the KDD99 data set,and inputs the results of the data preprocessing operation into the YOLOv5 detection algorithm.YOLOv5 uses the Mosaic data enhancement method to transform the data image Randomized splicing was performed to achieve data amplification.While enriching the attack data in the KDD99 data set,the network training speed is improved,and then rich information features are extracted from the input image,and finally the data is multi-scale detection to complete the fusion of different scale feature information,and accurately predict the data type.and classification for the purpose of intrusion detection.The model structure is relatively simple and easy to deploy,thereby improving the detection efficiency.