Intrusion Detection System Based On Radial Basis Neural Network

With the development of the wave of information revolution,the network has developed rapidly.The popularity of Internet and the number of Internet users have reached the peak of the same period in history.People can enjoy various conveniences such as online shopping,point takeaway,book hotels,and air tickets on various online platforms.While enjoying the great benefits of network development in clothing,food,housing,and transportation,people are inevitably facing various serious network security problems.In the early years,firewall technology as the main means of network security protection,because it can't resist the internal attacks of the network and the limitations of real-time detection,prompted the birth of intrusion detection technology.The intrusion detection system developed based on intrusion detection technology effectively solves the limitations of firewall technology in two aspects: real-time detection and internal attack identification.Nowadays,in the face of increasingly complex network environments,traditional intrusion detection technologies also have problems such as reduced identification types and reduced accuracy of recognition.In order to cope with these problems in the field of intrusion detection,the advantages of artificial neural network self-learning,the combination of intrusion detection and artificial neural network have become an important branch of network security research.Focusing on this important research direction in the field of network security,this paper studies the intrusion detection technology based on radial basis neural network,and designs and implements the intrusion detection system based on radial basis neural network.The main work of the thesis includes the following four aspects:(1)The current research status of network security situation and intrusion detection technology is introduced.The intrusion detection technology,intrusion detection system and artificial neural network are explored and summarized.At the same time,the basis functions,network structure and related learning algorithms of radial basis neural networks are systematically studied.According to the distribution of KDD CUP 99 dataset,the training data and test data were selected,and the selected data was preprocessed.(2)In order to solve the problem of poor stability of radial basis neural network based on gradient descent learning algorithm,starting from the aspect of the initial value range of the model parameters,the range of the initial value of the parameter basis function is selected by the outlier detection algorithm.Perform effective reductions and use the same data for experimental testing.Experiments show that compared with the traditional radial basis neural network based on gradient descent learning algorithm,the accuracy and stability of detection are improved.(3)In order to make the training data have better training effect,improve the accuracy of intrusion detection,the error rate and the missed report rate are reduced,and the training data of the radial basis neural network model based on the k-means clustering learning algorithm is subjected to the outlier detection,and the experiment test is carried out.The experiment shows that the accuracy is higher and the overall effect is better.(4)An intrusion detection system based on radial basis function neural network is designed and implemented,in which the background and front end of the system are based on Java and HTML,respectively.the system can use the radial basis function neural network model to detect the intrusion of the data set,and display the predicted type and the corresponding data set content on the final detection page together.