Theory And Implementation Of The Secure Database

In many data models of secure database, the MLR model is very successful. It introduced the concept of data-borrow successfully and solved the polyinstantiation problem. However, it also brought a potential security problem. In the data-borrow operation of this model, a low-level user can modify the view of high-level user. This paper presents an improved security model to solve the problem. The model extends the Smith-Winslett model and introduces the amendatory data-borrow operation. At the same time, the model redefines polyinstantiation integrity and referential integrity. It eliminates the ambiguous semantics, fuzzy query and proliferation of tuples due to updates. Then this paper studies the problem of inference control. It presents methods of detecting the inference channels. Also, it describes a two-phase inference control strategy. This strategy includes static upgrade of the seurity level and dynamic inference control. Static upgrade of the security level algorithm can guarantee controlling inference channels with less information loss. When the database is running, dynamic inference control algorithm can cut off the inference channels well without efficiency decline. At the same time, the paper studies covert channel control. It introduces the taxonomy, identifications and resolvents of covert channel briefly. With the advantages of the information flow method and covert flow tree method, this paper proposes a covert flow diagram method and gives steps of searching the diagram. According to the various theories, a multilevel secure DBMS is designed and developed. The result of our experiment shows that the theory and algorithms are correct.