Inferring approximated models for integration testing of component-based software

Recently, component-based software has earned widespread notice and acceptance as a method that facilitates the development of today´s large, complex and very critical systems by integrating prefabricated small pieces of software called components or COTS. Components integration becomes an essential stage in the component-based software development Lifecycle. Testing components after integration is an important activity. Due to the unavailability of source code of integrated components and due to the lack of component information, integration testing becomes more difficult and very complex task. Furthermore, the formal models of integrated components are not always available. Therefore, we need to extract/learn them from test observations. This paper proposed an improvement on the existing learning algorithm to infer a model of integrated components in a more efficient way in order to bring down the worst case time complexity. Preliminary result from case study demonstrated the applicability of our refined algorithm.