Building dependability arguments for software intensive systems

Abstract

A method is introduced for structuring and guiding the development of end-to-end dependability arguments. The goal is to establish high-level requirements of complex software-intensive systems, especially properties that cross-cut normal functional decomposition. The resulting argument documents and validates the justification of system-level claims by tracing them down to component-level substantiation, such as automatic code analysis or cryptographic proofs. The method is evaluated on case studies drawn from the Burr Proton Therapy Center, operating at Massachusetts General Hospital, and on the Pret a Voter cryptographic voting system, developed at the University of Newcastle.