Target-Specific Adaptations of Coupling-Based Software Reliability Testing

This article presents some approaches to software reliability testing supporting high coverage of component or (sub-)system interactions while enabling the selection of test cases according to target and scenario-specific criteria. On the one hand, in order to allow for reliability assessment, automatic test generation approaches must support the provision of stochastically independent and operationally representative test data. On the other hand, crucial sub-system interactions must be tested as intensely as possible, with particular concern for the even distribution of testing effort or for the prioritization of domain-critical data. Depending on such application-specific peculiarities, different multi-objective optimization problems are approached by novel genetic algorithms, successively applied to an interaction-intensive example in order to illustrate their practicality.