Uncertainty analysis in software reliability of component-based systems

Several architecture-based software reliability models were proposed in the past. These models compute the point estimate of system reliability by plugging point estimates of unknown parameters into the model. These models however, discard the uncertainty of the parameters, that is, they do not attempt to answer the question how parameter uncertainties propagate into overall system reliability. Therefore, the traditional way of estimating software reliability by plugging point estimates may not be appropriate. This thesis is focused on uncertainty analysis of architecture-based software reliability models. In particular, we present uncertainty analysis using the following methods: entropy, perturbation theory, method of moments and Monte Carlo simulation. The choice of the most appropriate method is based upon data requirements, reliability measures, accuracy of the solution and scalability criteria. Entropy and perturbation theory methods study uncertainty analysis of software operational profile, while the method of moments and Monte Carlo simulation enable us to study how the uncertainty of parameters propagates into the reliability estimate. Each method for uncertainty analysis is applied and analyzed on various case studies