Safety arguments, software and system reliability

The aim is to discuss the nature of safety arguments to consider the role of system and software reliability evaluation in these arguments, and to outline an approach to supporting the development of safety arguments. The author reviews some existing work addressing the problems of evaluating systems to high levels of reliability such as 10 ^-9 failures per hour using `black box´ testing. He also considers ways of achieving confidence beyond testable levels through the use of prior beliefs and discusses some approaches to achieving strong prior beliefs. He uses these possible approaches to illustrate a canonical form for representing (safety) arguments, and to outline the characteristics of a tool which he is constructing for safety argument management