Design And Implementation Of Network Intrusion Detection System Based On Deep Learning

With the rapid development of computers and the Internet,the scope of the network has gradually broken through the limitations of traditional computer networks and has begun to merge into the bits of people's lives.Examples of smart homes,smart medical care,etc.are all vivid portrayals of the Internet of Things,which is the result of the rapid development of the Internet and the integration of things around humans.With the gradual expansion of human dependence on the Internet,every bit of personal information life has been integrated into the Internet,which makes the Internet a huge information warehouse.While the Internet brings convenience to human life,it also leaves a lot of hidden dangers for users' information security.Large servers are attacked,users' private information is leaked,and so on.When this kind of cyber attack occurs in the medical,military,financial and other fields,the damage caused is particularly huge.Therefore,the development of an intrusion detection system with high accuracy and self-innovation is particularly important.The network intrusion detection system is divided into two parts based on the deep learning framework: one is the intrusion detection module based on the deep learning model,and the other is the information display and business processing module.For the intrusion detection module,the mainstream deep learning frameworks such as ANN,CNN,RNN are used,and the model training is carried out through the KDD cup99 data set.The trained model can display the performance of each trained model through the information display module,and the user Select the final model for testing.Similarly,users can also use existing data sets to retrain the detection module.The information display and business processing modules are based on the mainstream ssm framework,spring MVC performs business scheduling,and active MQ is used in the background for intrusion detection log backup.The information detected by the local detection module is published and subscribed in real time to the other two other than the local Make a backup on the log server.For the collection of detection data,Redis non-relational database caching can be used to cope with the surge of high concurrency data,which is very effective for a large number of DDOS and malicious scanning situations.The system has been successfully applied.And complete real-time network data intrusion detection with higher accuracy and lower false alarm rate,and can update itself according to the data provided by users.The entire system can run smoothly in a high-concurrency environment and perform real-time data backup.