An embedded system is a combination of hardware and software subsystems. Interaction between these two subsystems may lead to unexpected behavior when faults are present in either. An effective technique is required to detect the presence of such interaction faults in an embedded system. We propose a test data selection technique for interaction testing in the embedded system using hardware fault injection and mutation test criteria. The proposed technique simulates hardware faults as software faults and uses these to mutate the software component. The mutants so created are then used as a means to select test data that differentiates the original program from the mutants. An experimental evaluation of the proposed technique is also presented.