A novel approach for deriving interactions for combinatorial testing

Combinatorial testing focuses on identifying faults that arise due to interaction of values of a small number of input parameters. Also known as t-way testing, it reduces the size of test set by selecting a minimal set of test cases that cover all the possible t-way tuples. An optimal value of t (degree of interaction) for t-way testing for the system would maximize fault detection count in minimal number of test cases. However, identification of an optimal value of for t-way testing for the system remains an open issue. In this paper, we present an approach to identify the interactions that exist in the source code, thereby reducing the count of interactions to be tested. DD path graph is generated from the source code and interactions are identified using data flow techniques. Two case studies are also discussed in order to demonstrate our approach. Experimental results indicate that our approach significantly reduces the count of interactions to be tested without significant loss of fault detection capability. The approach is extensible to large sized structured programs.