Using Dynamic Sensitivity Analysis to Assess Testability

This paper discusses sensitivity analysis and its relationship to random black box testing. Sensitivity analysis estimates the impact that a programming fault at a particular location would have on the program's input/output behavior. Locations that are relatively "insensitive" to faults can render random black box testing unlikely to uncover programming faults. Therefore, sensitivity analysis gives new insight when interpreting random black box testing results. Although sensitivity analysis is computationally intensive, it requires no oracle and no human intervention.