The computation of a set of performance indicators of a computer-based system can be achieved through dependability analysis. Researchers have proposed several methods and tools that have an ability to give a prognosis for the failure of a computer-based system. These tools and methods are classified into three main approaches: model-based, data-driven, and experience-based prognostics. Wherever sufficient real data is available, the data-driven approach is appropriate, which can be transformed into behavior models using Hidden Markov Models, which fall in a subclass of Bayesian networks. In a Bayesian framework, the estimates of reliabilities of components of a computer-based system are updated using operational profile data as new information of reliability of one or more node becomes available for the identification of robustness of a system. In this paper, we show, using Bayesian Networks, how to update the reliability of individual components and the reliability of a whole computer-based system when the reliability of any component in the system changes. We use a running safety-critical computer-based system from a nuclear power plant as a case study.