Research Of Intrusion Detection System Based On Multiple-classifiers And Application On Meteorological Network Security

Current security of technical network is border defense and internal protection, such as firewall, safety routers, VPN equipment and so on. To prevent the system from illegal invasion they have certain effect, but the security technology can only plays a limited role of defense. Therefore, the traditional technology can not meet network security needs. Intrusion detection technology can effectively make up for the shortages. Intrusion detection system provides a good security for the network system. It is important to carry out the research of this aspect for improving information security level.The security of meteorological broadband network system based on the public network is sharing information and efficient operation of the important guarantee. It should guarantee the safety and reliability of the network. And then it ensures the confidentiality of the business application system completely. It provides the application space for expanding for the future. At present, the provinces departments for internal LAN provide meteorological business information exchange work and the structure of network is complex with the numerous and unsafe net sections. It is imminent to develop and implement meteorological network security system. While the traditional intrusion detection system based on single classifier performance not that good on heterogeneous data and diversification data, this paper put forward the intrusion detection system model based on multiple classifiers is applied to the network in order to adjust the balance of the false positive rate and false negative rate.The main work of the paper concludes the following:(1)It expounds the support vector machine classifier and application on the intrusion detection. A variety SVM method of feature extraction for the invasion of the classification is used after data processing. Designing the system of multiple-classifiers, the paper shows the intrusion detection experiment results, adopting KDD CUP99data set.(2)An intrusion detection is designed which divides into network acquisition module, the data pretreatment module, the classifier with feature extraction modules and response module, where each part is discussed in detail. According to the problem of network security, multiple-classifiers intrusion detection system model is used that proves the superiority. And then it introduces the other security linkage mechanisms (taking firewall for example) for solving the problem of network security and the deployment of the method. According to the business systems in the weather bureaus, the deployment proposals for security of network are put forward.The main innovation or improvement of this thesis is: (1)Various feature extractions and combination classifiers are used to improve generalization capability of intrusion detection. The regulators of classifier weight coefficients are done to balance the false positive rate and false negative rate.(2)The intrusion detection system model based on multiple classifiers is applied to the construction of the meteorological network security, combining with other network security technology for linkage response. The solution of both security technology and safety management is put forward.