Évaluation des logiciels critiques: application à la méthode B

Abstract

From the point of view of safety the bottleneck in the operation of computerized systems currently appears to be the software. In order to improve the safety of software an independent assessment of its quality has been advocated for several years. The assessment of safety critical software consists of a technical examination on the basis of a set of criteria by a body which is independent from the designer. These assessment criteria are derived from the standards which are in force, the state of the art and experience. The objective of an assessments to verify that the product meets the specified safety and functionality requirements. In this paper the authors describe a combined approach to the assessment of safety critical software. On the one hand, the process used in order to develop the software is examined, and on the other the quality of the end product (software) is checked. The first part of the paper will present the position of safety critical software in the context of railway standardization as well as the main assessment techniques with particular reference to that used at Inrets. The second part of the paper will deal with the contribution of formal methods such as the B method to the development of safety critical software. The software assessment aspect will pay particular attention to assessment of the product — analysis of the traceability of requirements, integration and validation of safety properties. The paper ends with a brief account of the extraction and interpretation of metrics for a formal development. Finally the criteria used to assess the process are rapidly described. This approach has been used in the framework of an Esprit project known as Cascade (Certification and Assessment of Safety Critical Application Development).