Defining Abstract Semantics for Static Dependence Analysis of Relational Database Applications

Dependence Graph provides the basis for powerful programming tools to address a large number of software engineering activities including security analysis. This paper proposes a semantics-based static dependence analysis framework for relational database applications based on the Abstract Interpretation theory. As database attributes differ from traditional imperative language variables, we define abstract semantics of database applications in relational abstract domain. This allows to identify statically various parts of database information (in abstract form) possibly used or defined by database statements, leading to a more precise dependence analysis. This way the semantics-based dependence computation improves w.r.t. its syntax-based counterpart. We prove the soundness of our proposed approach which guarantees that non-overlapping of the defined-part by one statement and the used-part by another statement in abstract domain always indicates a non-dependency in practice. Furthermore, the abstract semantics as a basis of the proposed framework makes it more powerful to solve undecidable scenario when initial database state is completely unknown.