| As an important part of corporation information system, database stores a great deal of data, among which the security of data plays an crucial part in securing the safety of the information system. At present, it needs to improve the security control of the database itself. The data information may be amended or filched by vicious user and canting super visitor from the inner side or outside. In order to maintain the security of data information, it is insufficient to depend on the security system of database itself only.The thesis combines the method of neural networks capacity with the technique of intrusion detection, studying the security techniques of protecting corporation database, and developing a corresponding auxiliary safe database system. On the base of intrinsic network patulous structure and safe mechanism in corporation, according to the universal framework of intrusion detection, the thesis constructs a model of intrusion detection system applied to the corporation database. With combination of detection technique of abnormity and misuse in intrusion detection, the model can detect unknown and known intrusions.The thesis studies the security mechanism of the corporation database SQL Server, and obtains the data source by tracing client's transfer using SQL Server event profile tool. The filtered and pretreated data can reflect the transfer characters of users in conformation to the input requirement of neural network; the thesis further studies the way to improve BP arithmetic neurotic network generalization by adding dynamic item, changing study rate and introducing the sum progression measures; it also contrast the improved arithmetic method with traditional method. The improved method can effectively improve the generalization ability of neurotic network and the convergent rate. The thesis also discusses the structure implement of BP network, applying the BP arithmetic to intrusion detection. Based on the intrusion detection model, the thesis designs and actualizes a host-based enterprise database intrusion detection system on windows platform, which can achieve online detection and offline analysis, combining initiative respond technique with passiveness respond. The system can ensure data integrity and veracity, as well as afford orbicular information with intrusion detection audit.The thesis does a research in one kind of database security protection technology within the corporation by using modified BP neural networks, which can effectively prevent offside operation and vicious destroy from inside, and rejects long-distance lawless intrusions from outside. It is an effective attempt to solve the problem of safety within the corporation MIS database by using the neural networks technique. |
