Challenges in Assuring Highly Complex, High Volume Safety-Critical Software

Abstract

Safety standards such as IEC 61508 [1] and ISO 26262 [2] were written assuming that highly complex safety-critical systems were self-contained, capital intensive, one-off products and that the assurance argument was to be renewed with every product change. In this paper, we explain how the increasing complexity, scale, continuous updates and heterogeneous nature of future safety-critical software systems requires a paradigm shift in the software safety qualification and certification processes. We examine existing approaches to software safety argumentation and explain how the essential components of the argumentation can be transferred to future software architectures and development processes. In particular, we discuss the relevance of constructive measures for ensuring safety, evidence collection to demonstrate the effectiveness and integrity of these measures and process approaches to assess the predictability of the overall quality and integrity of the software. Our aim is to provide an industrial perspective on the areas of collaboration required to transfer relevant research results into future standards.