Eliciting system requirements is not an easy task, especially if one considers the amount of problems that can arise once the system is at use. A difficulty to ensure requirements completeness is to foresee these problems and identify the requirements that can help to overcome them. Existing approaches are based on probabilistic prediction of usage or system-related problems. Very few of these approaches are able to take into account both the causes of the potential problems and the impacts that these problems can have on the system usage. This paper presents an approach that proposes to elicit requirements by analysing how such kinds of problems can threaten the adequacy of a system to its intended usage. The approach aims at improving requirements completeness by (i) guiding a systematic identification of threats, (ii) evaluating their cause and (Hi) evaluating their consequences on the system. This approach extends the CREWS -l'Ecritoire method with techniques adapted from the risk engineering domain. CREWS-l'Ecritoire couples abstract goals with concrete scenarios to elicit requirements. It proposes a number of rules to model goals, author scenarios, and discovers new goals by analysing scenarios. Our proposal stands in a new kind of goal discovery rule that is based on threat discovery, cause and consequence analysis, and solution identification. The paper gives an overview of the proposed rule, then shows how it integrates into CREWS-l'Ecritoire, and provides an example of application.
展开▼