Change Sensitivity Based Prioritization for Audit Testing of Webservice Compositions

Modern software systems have often the form of Web service compositions. They take advantage of the availability of a variety of external Web services to provide rich and complex functionalities, obtained as the integration of external services. However, Web services change at a fast pace and while syntactic changes are easily detected as interface incompatibilities, other more subtle changes are harder to detect and may give raise to faults. They occur when the interface is compatible with the composition, but the semantics of the service response has changed. This typically involves undocumented or implicit aspects of the service interface. Audit testing of services is the process by which the service integrator makes sure that the service composition continues to work properly with the new versions of the integrated services. Audit testing of services is conducted under strict (sometimes extreme) time and budget constraints. Hence, prioritizing the audit test cases so as to execute the most important ones first becomes of fundamental importance. We propose a test case prioritization method specifically tailored for audit testing of services. Our method is based on the idea that the most important test cases are those that have the highest sensitivity to changes injected into the service responses (mutations). In particular, we consider only changes that do not violate the explicit contract with the service (i.e., the WSDL), but may violate the implicit assumptions made by the service integrator.