Model-based integration testing technique using formal finite state behavioral models for component-based software

Abstract

Many issues and challenges could be identified when considering integration testing of Component-Based Software Systems (CBSS). Consequently, several research have appeared in the literature, aimed at facilitating the integration testing of CBSS. Unfortunately, they suffer from a number of drawbacks and limitations such as difficulty of understanding and describing the behavior of integrated components, lack of effective formalism for test information, difficulty of analyzing and validating the integrated components, and exposing the components implementation by providing semi-formal models. Hence, these problems have made it in effective to test today’s modern complex CBSS. To address these problems, a model-based approach such as Model-Based Testing (MBT) tends to be a suitable mechanism and could be a potential solution to be applied in the context of integration testing of CBSS. Accordingly, this thesis presents a model-based integration testing technique for CBSS. Firstly, a method to extract the formal finite state behavioral models of integrated software components using Mealy machine models was developed. The extracted formal models were used to detect faulty interactions (integration bugs) or compositional problems between integrated components in the system. Based on the experimental results, the proposed method had significant impact in reducing the number of output queries required to extract the formal models of integrated software components and its performance was 50% better compared to the existing methods. Secondly, based on the extracted formal models, an effective model-based integration testing technique (MITT) for CBSS was developed. Finally, the effectiveness of the MITT was demonstrated by employing it in the air gourmet and elevator case studies, using three evaluation parameters. The experimental results showed that the MITT was effective and outperformed Shahbaz technique on the air gourmet and elevator case studies. In terms of learned components for air gourmet and elevator case studies respectively, the MITT results were better by 98.14% and 100%, output queries based on performance were 42.13% and 25.01%, and error detection capabilities were 70.62% and 75% for each of the case study.