A validation metrics framework for safety-critical software-intensive systems

Abstract

Validation of safety-critical software requirements is a difficult and frequently misunderstood task. It answers the question of "are we building the right product?" and is essential to Software Engineering. However, validation is often confused with verification activities, or simply left as a final tick-in-the-box just prior to delivery. Current models for validation cannot satisfy the unique aspects of safety-critical software where "building the right safety product" is paramount. Software safety requires a new model for validation of safety requirements by proxy. The need for a proxy model becomes evident in the software safety process, where customer input for safety is reduced to the requirement of "a safe system." This thesis defines a new proactive model for validation of safet-critical software requirements. Continuous assessment of validity of safety requirements is indicated by metrics as part of the Validation Metrics Framework. The generic framework combines the Goal/Question/Metric Approach with Goal Structuring Notation and then specializes in validation of safety-critical software. The metrics are measurements of safety products typical to safety-critical software development programs. A fictitious case study of a Rapid Action Surface to Air Missile is used to apply the framework, identifying the benefits of a proactive, indicative, validation technique utilizing a metrics framework.