A BBN approach to certifying the reliability of COTS software systems

Best practices indicate that totally independent certification is the only safe approach to take when certifying a system assembled from COTS software components. We explore one method of certifying the reliability of COTS software systems in this paper. Certifying the reliability of a COTS software system can be problematic, given that we may have incomplete knowledge of a component and its ability to interact in a reliable manner with other components; many times, all we have to evaluate is the behavior of the components as described by their vendor(s) and evidence gathered during testing. We offer a Bayesian Belief Network (BBN) that will help us resolve what we believe to be one of the major areas of uncertainty in certifying the reliability of a COTS software system and develop a reasoned level of belief in the reliability of the system in this work. Uncertainty is inherent and inevitable in software development processes and products. Active research is being performed in COTS software risk management, and several achievements have been published. However, uncertainties in COTS software remain largely unexplored, and this problem has caught researchers´ attention since no one can use a system without the certain certified reliability. Our overall goal is to develop a systematic framework to interpret the sources of uncertainties inherent in COTS software using BBNs. A set of BBNs is designed to represent different types of uncertainty in the COTS software certification process. We use the object oriented BBN methodology to frame all the BBNs related to the set of uncertainties. The BBN will provide a quantitative measure of the system´s reliability.