Second-Order Mutation Testing Cost Reduction Based on Mutant Clustering using SOM Neural Network Model

Abstract

Second-order mutation testing aims to simulate more actual and more complicated program defects by manually injecting two errors into the original programs, which is indispensable and significant in current mutation testing related studies. However, exponential growth of the number of second-order mutants and subtle second-order mutants detection are still major obstacles. In this paper, we propose a novel second-order mutants reduction method which well utilizes the Self-organizing Maps (SOM) neutral network model based mutants clustering tactic. First, a better combination strategy is used to generate feasible second-order mutants based on traditional first-order mutant generation, and then we construct accurate SOM neural network model according to the similarity of intermediate values in the execution of second-order mutants, and at last mutants are clustering based on such model to achieve second-order mutant reduction and subtle mutants detection. Through analysis of experiment results, our method could greatly reduce the number of second-order mutants without big loss of mutation scores, and correspondingly reduces the execution costs for second-order mutation testing. Besides, our method could effectively find subtle second-order mutants when considering the combination of any two faults in the original program.