Formal Modeling of Distributed Architectures with FreeSpec: A Cloud-Based MTD Use Case

This paper examines the formal modeling of distributed software architectures deployed in untrusted cloud environments. We present a use case from medical imaging, where a learning system must preserve the confidentiality of both training data and model outputs. The system integrates a moving target defense (MTD) strategy to mitigate the risk of inference by a cloud-based adversary. We leverage FreeSpec, a framework for specifying software via compositional interfaces, and use Rocq (formerly Coq) to reason about the system’s information flow properties. Our analysis demonstrates how FreeSpec can be used to capture such architectures, while revealing two key limitations: the unidirectional nature of component interactions, and constraints in the composition model based on free monads. The work outlines a methodology for applying formal reasoning to cloud-based MTD systems, and identifies requirements for extending current frameworks to address distributed security challenges.