Domain Separation by Construction

Language-based approaches to security typically use static type systems to control information flow, relying on type inference to distinguish secure programs from insecure ones. This paper advocates a novel approach to language-based security: by structuring software with monads (a form of abstract data type for effects), we are able to maintain separation of effects by construction. The thesis of this work is that well-understood properties of monads and monad transformers aid in the construction and verification of secure software. We introduce a formulation of non-interference based on monads (rather than the typical trace-based formulation). Using this formulation, we prove a noninterference style property for a simple instance of our abstract system model. Because monads may be easily and safely represented within any higher-order, typed functional language, monadic event systems may be directly realized within such a language.