Citation

Hawkins R, Habli I, Kelly T, McDermid J. Safety Sci. 2013; 59: 55-71.

Copyright

(Copyright © 2013, Elsevier Publishing)

DOI

10.1016/j.ssci.2013.04.007

PMID

unavailable

Abstract

In safety-critical applications, it is necessary to justify, prior to deployment, why software behaviour is to be trusted. This is normally referred to as software safety assurance. Within certification standards, developers demonstrate this by appealing to the satisfaction of objectives that the safety assurance standards require for compliance. In some standards the objectives can be very detailed in nature, prescribing specific processes and techniques that must be followed. This approach to certification is often described as prescriptive or process-based certification. Other standards set out much more high-level objectives and are less prescriptive about the particular processes and techniques to be used. These standards instead explicitly require the submission of an assurance argument which communicates how evidence, generated during development (for example from testing, analysis and review) satisfies claims concerning the safety of the software. There has been much debate surrounding the relative merits of prescriptive and safety assurance argument approaches to certification. In many ways this debate can lead to confusion. There can in fact be seen to be a role for both approaches in a successful software assurance regime. In this paper, we provide a comparative examination of these two approaches, and seek to identify the relative merits of each. We first introduce the concepts of assurance cases and prescriptive software assurance. We describe how an assurance case could be generated for the software of an aircraft wheel braking system. We then describe how prescriptive certification guidelines could be used in order to gain assurance in the same system. Finally, we compare the results of the two approaches and explain how these approaches may complement each other. This comparison highlights the crucial role that an assurance argument can play in explaining and justifying how the software evidence supports the assurance argument, even when a prescriptive safety standard is being followed.