Metrics design for safety assessment

Context:In the safety domain, safety assessment is used to show that safety-critical systems meet the required safety objectives. This process is also referred to as safety assurance and certification. During this procedure, safety standards are used as development guidelines to keep the risk at an acceptable level. Safety-critical systems can be assessed according to those safety standards. Objective:Due to the manual work, safety assessment processes are costly, time consuming, and hard to be estimated. The goal of this paper is to design metrics for safety assessment. These metrics can, for instance, identify costly processes in the safety assessment process. In this paper we propose a methodology to design metrics for safety assessment from different perspectives. For the demonstration and validation of our method, we focus on safety assessment in the automotive domain (ISO 26262). Method:Metrics can be identified by answering three questions. Three different sources of information have been identified for obtaining metrics: industrial interests, safety standards, and available data. For each of these sources appropriate methods have been proposed and used for obtaining the relevant metrics. These methods include GQM-based surveys, PSM-based procedure, and brainstorming. For the validation, the ISO 26262 standard has been studied for obtaining safety standard related metrics. Results:A case study in the context of the European project OPENCOSS is carried out to demonstrate the method. Finally, there are 76 metrics obtained and a validation of these metrics has been done by means of a survey amongst 24 experts from 13 project partners. Conclusion:It can be concluded that metrics for safety assessment can be derived from three sources. Different methods for designing metrics have to be used for each source. The validation shows that most of the relevant metrics are useful for industry.