The Research On Metamorphic Relation Sensitivity

Software Testing is an important way to guarantee and improve the software quality, and it aims to detect as many bugs as possible in the software. However, a fundamental problem of the software testing is oracle problem which, in many situation, is not easy for tester to get expert result for a given program input, and hence, is hard to get test result. Metamorphic testing(MT) is proposed to alleviate the oracle problem. It verifies the test results by comparing the original test set and the follow-up test set to satisfy the metamorphic relation(MR).MR plays a key role in MT, which not only helps to generate follow-up test cases,but also directly influences the test results and test efficiency. Hence, One core study topic of the MT is the selection good MR. In the existing paper, mutation score(MS)or fault detection rate(FDR) is a popular way to measure the fault detection ability of MR. MS covers some important characteristics of the MR that is MR maintaining different fault detection ability for different faults. And FDR can’t overcome one-sidedness and blindness. Hence, to avoid the disadvantage of the existing measure ways, this paper proposes a new concept called metamorphic relation sensitivity(MRS). MRS can reflect more fully characteristics of MR.MRS is a multi-dimensional feature vector which can be distributed on the FDR for various kinds of possible faults of the software under testing(SUT). It depicts the sensitive degree of the MR for different faults and gives a measurable method for connecting the MR with the faults of SUT, and provides a new perspective and possibility for the study of the MT.Due to the diversity fault of the SUT, the dimension of the MRS is really high. To reduce the difficulty of calculation MRS, we propose a new strategy based on the mutation subset to reduce the dimension of the MRS.The MRS is widely used in the study of software testing. First of all, MRS can be used for the cluster analysis on MR set. The experiment results show that the MRS supports cluster analysis strongly and offer some useful knowledge. Another important application of the MRS is analysis the composition metamorphic relation(CMR). The experiment results can get conclusion that the hybrid metamorphic relation(HMR) has more advantage on constructing CMR than that self-metamorphic relation(SMR) and the Q property of CMR is existing. The last application of the MRS is team metamorphic relation(TMR). And experiment results prove that the team metamorphic testing(TMT) has a better test result, and we also study what factors affect the fault detection rate of overlapping and proposal a algorithm to construct the better TMR.