Confidence in safety argument - An assessment framework based on belief function theory. (Confiance dans un argumentaire de sécurité - un cadre d'évaluation basé sur la théorie des fonctions de croyance)

Rui Wang

Abstract: Safety arguments, also called Safety Cases, are commonly used to present that adequate efforts have been made to achieve the safety goals. Thus, the system safety is often justified through assessing the safety arguments. The assessment of such arguments is usually implemented by experts without any dedicated tool or method. This leads to a questionable validity of the results. In this thesis, a quantitative framework is proposed based on Dempster-Shafer theory (D-S theory) to assess our confidence in Safety Cases. This framework manages the issue in following aspects: 1) Formal definition of confidence in arguments based on D-S theory; 2) Development of confidence aggregation rules; 3) Proposition of a quantitative assessment framework of safety arguments. An application in railway domain realises the parameter estimation of the framework by a survey with safety experts.