Improving system reliability assessment of safety-critical systems using machine learning optimization techniques

Abstract

Quality assurance of modern-day safety-critical systems is continually facing new challenges with the increase in both the level of functionality they provide and their degree of interaction with their environment. We propose a novel selection method for black-box regression testing on the basis of machine learning techniques for increasing testing efficiency. Risk-aware selection decisions are performed on the basis of reliability estimations calculated during an online training session. In this way, significant reductions in testing time can be achieved in industrial projects without uncontrolled reduction in the quality of the regression test for assessing the actual system reliability.