Execution Optimization Technology For Mutation Based Fault Localization

Mutation Based Fault Localization(MBFL)is an automatic fault localization technique based on mutation testing,which has the advantage of high fault localization accuracy.On the other hand,it needs to generate a large number of mutants,and each mutant has to be executed on all test cases.So the execution cost of MBFL is extremely high,which severely restricts its practical application in the industrial field.Existing MBFL optimization techniques mainly consider the aspects of the mutant sampling and execution process,and these techniques are suffered from low mutation reduction rate or loss of fault localization accuracy.Based on the analysis of the execution process of MBFL,this paper considers the execution optimization method from two aspects,which are mutant reduction and test case reduction.It aims to implement an MBFL technique with low mutation execution cost and high fault localization accuracy.1.In terms of mutant reduction,this paper proposes a fuzzy domination based mutant optimization strategy(FDMOS)by studying the relationship between dominant mutants and fault localization accuracy.This method selects a small number of dominant mutants for fault localization,which significantly reduces the execution cost of MBFL under the premise of keeping high fault localization accuracy.2.In terms of test case reduction,this paper proposes an information entropy based test case reduction strategy(IETCR).IETCR first calculates the information entropy of test cases,then sorts them according to the information entropy and finally it selects a small number of valuable test cases to execute mutants.IETCR achieves the goal of improving the execution efficiency of MBFL by reducing the size of test cases.Since IETCR can keep test cases with higher impacts,so using this technique in MBFL can achieve high fault localization accuracy.3.Based on the above two methods,this paper proposes a hybrid approach by combining mutant reduction and test case reduction(MRS)to reduce the execution cost of MBFL further.In order to evaluate the effectiveness of the proposed methods,this paper conducted empirical studies on 112 fault versions of 6 programs from the SIR repository.The experimental results show that using FDMOS,IETCR,and MRS can reduce about 41.2%?81.6%,56.3%?88.3%,and 78.2%?95.7%mutation execution cost of MBFL,respectively.Furthermore,this paper employs a Wilcoxon signed-rank test to statistically analyze the fault localization performance of different techniques,and the results indicate that there is no significant difference between MBFL with our proposed methods and the original MBFL.