Concept of a Value in Multilevel Security Databases

In this paper, we focus on a framework of multilevel security databases. We assume subjects form a hierarchy and they have varying beliefs about existence, identities, and other properties of objects. Every subject has an object space that represents its own belief about the real world. A subject is interested in not only query of objects in its own space, but also how they relate to objects belonging to lower subjects. We introduce the concept of a skeleton that allows all cascading beliefs of a given object to be assembled in a way that is natural for users in multilevel security. Skeletons are treated as most primitive values in the model. In stored relations, tuples consist of skeletons, one for each attribute of the tuple. Any query related to this object would draw a subset of the tuple. Skeletons greatly simplifies query of cascading beliefs, the main preoccupation of a subject in multilevel security. Self-joins are largely reduced. Query is much more simplified compared to other models. We also show that the skeletons obey good algebraic identities which are a prerequisite for an SQL-like declarative algebraic query language. Depending upon their thought process, different users can formulate a query in variety of ways. The formalism of the model presented in this paper proves to reassure the users that queries that seem equivalent are indeed equivalent and therefore retrieve the same result. The operations are used freely without any fear of loss of information. Intricate details of algebraic operators introduced in this paper are meant to be absorbed by the system and not passed on to the users. Thus this paper lays stronger foundation for natural query of multilevel security data than existing works.