Composition and Declassification in Possibilistic Information Flow Security

Abstract

Formal methods for security can rule out whole classes of security vulnerabilities, but applying them in practice remains challenging. This thesis develops formal verification techniques for information flow security that combine the expressivity and scalability strengths of existing frameworks. It builds upon Bounded Deducibility (BD) Security, which allows specifying and verifying fine-grained policies about what information may flow when to whom. Our main technical result is a compositionality theorem for BD Security, providing scalability by allowing us to verify security properties of a large system by verifying smaller components. Its practical utility is illustrated by a case study of verifying confidentiality properties of a distributed social media platform. Moreover, we discuss its use for the modular development of secure workflow systems, and for the security-preserving enforcement of safety and security properties other than information flow control.