A Novel Approach to Measure Confidence and Uncertainty in Assurance Cases

Abstract

Assurance cases are a well-established structured technique used to document a reasoned, auditable argument supporting that a system meets desirable properties (e.g., safety or security). Assurance cases are increasingly becoming popular, and are being used to make safety and cyber-security arguments about medical, automotive and aviation systems. Current methods usually assess confidence in assurance cases, but only with evidence available at design-time. However, real-world situations demand considerations of evidence that are also available at run-time. In this paper, we introduce a novel confidence measure called INCIDENCE (weIghted assuraNCe confIDENCE). The measure considers evidence available both at design and run times, and is suitable for the assessment of assurance cases represented using Goal Structuring Notation (GSN) – being a popular notation for representing assurance cases. We rely on the confidence measure to derive an uncertainty measure that can be used to measure technical debt (requirement debt) for software systems. We illustrate our work through an example focusing on feature identification.