Towards A Formal And Scalable Approach For Quantifying Software Reliability At Early Development Stages
MetadataShow full item record
Problems which originate in early development stages can have a lasting influence on the reliability, safety, and cost of a software system. The requirements document, which is usually available at the requirements analysis stage, must be correct, unambiguous, and complete if the rest of the development effort is to succeed. The ability to identify faults in requirements and predict the reliability of a software system early in its development can help organizations make informative decisions about corrective actions and improve the system's quality in a cost-effective manner. A review of the literature reveals that existing approaches are unsuited to provide trustworthy reliability prediction either due to the ignorance of the requirements documents, or because of the informal and fairly sketchy way in detecting faults in requirements. This study explores the use of a preselected software reliability measurement for early software faults detection and reliability prediction. This measurement, originally a black-box testing technique, was broadly recognized for its ability to detect incomplete and ambiguous requirements, although no information was found in the literature about how to take advantage of its power. This study mathematically formalized the measurement to enhance its rigidity, repeatability and scalability and further extended it as an effective requirements faults detection technique. An automation-oriented algorithm was developed for quantifying the impact of the detected requirements faults on software reliability. The feasibility and scalability of the proposed approach for early faults detection and reliability prediction were examined using two real applications. The results clearly confirmed its feasibility and usefulness, particularly when no failure data is available and other methods are not applicable. The scalability barriers were also spotted in the approach. An empirical study was thus conducted to gain insight into the nature of the technical barriers. As an attempt to overcome the barrier, a set of rules was proposed based on the observed patterns. Finally, a preliminarily controlled experiment was conducted to evaluate the usability of the proposed rules. This study will enable software project stakeholders to effectively detect requirements faults and assess the quality of requirements early in development, and ultimately lead to improved software reliability if the identified faults are removed in time. Software project practitioners, regulators, and policy makers involved in the certification of software systems can benefit most from the techniques proposed in this study.