Research On Intrusion Detection Based On Neural Network

With the rapid development of the Internet,network security incidents have emerged in an endless stream,which has brought immeasurable losses to the economies of various countries and has also seriously hindered the further development of network technologies.Traditional network security defense technologies can only provide static protection,such as access control,encryption and decryption technologies,rule-based anti-virus technologies,and firewalls.Intrusion detection systems are an important complement to traditional technologies.They can be acquired and analyzed in real time through operations.The event information of the system,network and user can achieve the purpose of detecting intrusion events without affecting the normal use of the network.However,existing intrusion detection systems have low detection performance and face more and more complex network environments,and researches are more efficient and perform better.High intrusion detection systems have important practical significance.This paper firstly analyzes the current computer network security situation and the purpose and significance of the intrusion detection system.It elaborates the research status of the intrusion detection system at home and abroad,and introduces the relevant theories and techniques of intrusion detection,and summarizes the machine learning-based Intrusion detection model general architecture and performance evaluation indicators.For the current intrusion detection technology detection accuracy is generally not high,training time is too long and other shortcomings put forward some optimization measures.Firstly,a novel feature selection method based on Fisher-PCA is proposed for the problem of high-dimensional network data that will lead to dimensional disasters.It can effectively reduce the training time of the model,solve the problem of low efficiency of model training,and then aim at the traditional BP neural network.The network convergence speed is slow and it is easy to fall into the local optimal value.A neural network model based on Adam adaptive algorithm optimization is proposed.Combining feature selection method and Adam-DNN design,an intrusion detection model is implemented.The data is used as training data of the model and experimentally simulated.The results show that this method effectively improves the training efficiency and detection accuracy of the intrusion detection system.