Didkovska M. Estimation methods and means of software reliability increase

The thesis is devoted to the development of software reliability estimation models and means of software reliability increase during the whole software life cycle, according to the modern tendencies in software development. Generalized and extended software reliability estimation model based on the nonhomogeneous Poisson process was developed. Function of failure intensity of this model has complex form, which coincides with the results obtained in practice. It allows to take into account the process of learning during testing. The application of reliability grows models is limited, because it's not always possible to obtain all necessary input data for models. In such a case a testing can be used. During the investigation the new integration testing criteria and measures that are based on design description of software component interaction were introduced. They are innovative and utilize formal design description as a basis. These criteria have practical value because the fulfillment of criteria canbe automatically checked using well-known notation of UML-diagrams. Formal coverage criteria offer testers ways to select test data and provide necessary stopping rules. They allow to develop test plans in the early phases of software development.