Software Safety

Abstract This is an age of rapidly increasing technological dependence on computer‐controlled systems. Given the relative and obvious unreliability of the humble but ubiquitous PC , the reader may wonder just how safe this trend might be. In this article, the author try to answer this question. The safety of any system is reflected in its inability to do harm. A system is unsafe if harm can result from its operation and we then call this a safety‐critical system. If a system is safe only within certain operating bounds, then these must be explored and defined. Assessing the safety of a system can be a very complex process involving many factors including but not limited to reliability, design resilience, human interaction, and interactions with other systems. As systems evolve into ever more complex forms, this just gets harder. From this we can see that safety is a property of a system as a whole . In this article, only systems are addressed. By definition, any computer‐controlled system contains software of one kind or another, and computer‐controlled systems and software‐controlled systems are treated as synonymous. The software contained therein might be embedded in the system in the form of firmware or may be readily accessible. From the point of view of safety, however, there is no effective distinction. Software‐controlled systems are essentially treated in the same way as conventional engineering systems such as bridges, but the scale of software‐controlled systems is now much larger and many of the associated risks turn out to be essentially unquantifiable at the current state of knowledge. Computer‐controlled systems can take many forms. In some systems, software may contribute only a small part to the overall operation and the failure of the software component may not prejudice the overall behaviour at all. In such systems, we say that software is a noncritical component. In contrast, for other systems, software may contribute a major part or even the entire behavior of the system. Failure of the software in such a system prejudices the overall behavior of the system, perhaps catastrophically. In these systems, we say that software is a critical component. It is getting more and more difficult to assess the safety or reliability of software controlled systems. There are essentially five contributing factors: Size, Complexity, Coupling, Ubiquity, and Sensitivity. The influence of design and implementation and standardization initiatives are discussed.