Automated Method for Assurance Case Construction from System Design Models

Abstract

Many Cyber Physical Systems (CPSs) operate in safety- or mission-critical applications where strong assurance of safe operation is required. Assurance cases are one widely used tool for presenting an argument in support of system safety requirements backed by a body of evidence - e.g., test data, formal analysis, or expert review. A system assurance case should regularly evolve during the development process as the system design is refined, but construction and refinement of the assurance argument is a labor-intensive and error-prone process. Reuse of common patterns from successful arguments is one prominent technique for reducing the required effort, but instantiation of these patterns still requires manual compilation and organization of the relevant information from system artifacts. Since model-based engineering techniques are common for CPS development, the relevant information is often contained in a set of interconnected models describing the system. In this paper, we present a method for assurance case construction based on the instantiation and composition of patterns. This method automates the collection and organization of necessary information by extracting it directly from an existing set of system design models. To support human review and refinement, the generated assurance case maintains (1) traceability from objects in the argument back to the corresponding system models and (2) explain ability of choices made during the construction process based on the relationships between model objects. An illustrative example is provided which demonstrates the utility of this method to produce assurance cases which are tightly integrated with source artifacts while requiring significantly less effort than the manual construction alternative.