Specifying Safety Requirements with GORE languages

Context: A suitable representation of Safety-Critical Systems (SCS) requirements is crucial to avoid misunderstandings in safety requirements and issues in safety specification. However, current general requirements specification languages do not fully support the particularities of specifying SCS. Objective: In this paper, our goal is to identify and propose a set of important features that should be provided by requirements languages to support an early safety requirements specification. Moreover, we aim to compare the ability of the four most used Goal-Oriented Requirements Engineering (GORE) languages (i*, KAOS, GRL, NFR-Framework) in supporting the proposed features. Method: We first established a conceptual foundation and a conceptual model based on the literature, challenges elicited in previous works, and demands of safety standards at the requirements level that practitioners must satisfy in order to certify their systems. Results: We proposed a set of 15 features that requirements languages should provide to an early safety requirements specification. Regarding the comparison of GORE languages, in summary, all surveyed languages lacks explicit modeling constructs to express how hazards can occur in the system, the accidents, their impact and how they can mitigated. Conclusions: The conceptual foundation, conceptual model, and the set of features is a novelty. Finally, the features can be used to propose new requirements languages for SCS or to define extensions for the ones already available.