SCOLL -- A Language for Safe Capability Based Collaboration

In capability secure systems it is important to understand the restrictive influence programmed entities (e.g. procedures, objects, modules, components) have on the propagation of influence in a program. We explain why Take-Grant systems are not sufficiently expressive for this task, and we provide a new formalism – Authority Reduction systems (AR-systems) – to model collaborative propagation. AR-systems provide safe and tractable approximations of adequate precision for the confinement properties in configurations of collaborating entities. We propose a domain specific declarative language – SCOLL (Safe COLlaboration Language) – to express the collaborative behavior of subjects, the initial conditions in a configuration, and the requirements about confinement and liveness that are to be ensured. We provide the syntactic structure and an operational and denotational semantics for the language. From experiments with a first implementation, we provide a preliminary result and show how patterns for capability based collaboration can be analyzed and generated