Experimental validation of a risk assessment method

Abstract

[Context and motivation] It is desirable that requirement engineering methods are reliable, that is, that methods can be repeated with the same results. Risk assessments methods, however, often have low reliability when they identify risk mitigations for a system based on expert judgement. [Question/problem] Our goal is to assess the reliability of an availability risk assessment method for telecom infrastructures, and to identify possibilities for improvement of its reliability. [Principal ideas/results] We propose an experimental validation of reliability, and report on its application. We give a detailed analysis of sources of variation, explain how we controlled them and validated their mitigations, and motivate the statistical procedure used to analyse the outcome. [Contribution] Our results can be used to improve the reliability of risk assessment methods. Our approach to validating reliability can be useful for the assessment of the reliability of other methods.