Refactoring Preserves Security

Refactoring allows changing a program without changing its behaviour from an observer’s point of view. To what extent does this invariant of behaviour also preserve security? We show that a program remains secure under refactoring. As a foundation, we use the Decentralized Label Model (DLM) for specifying secure information flows of programs and transition system models for their observable behaviour. On this basis, we provide a bisimulation based formal definition of refactoring and show its correspondence to the formal notion of information flow security (noninterference). This permits us to show security of refactoring patterns that have already been practically explored.