A source-code aware method for software mutation testing using artificial bee colony algorithm

Abstract

The effectiveness of software test data relates to the number of found faults by the test data. Software mutation test is used to evaluate the effectiveness of the software test methods and is one of the challenging fields of software engineering. In order to evaluate the capability of test data in finding the program faults, some syntactical changes are made in the program source code to cause faulty program; then, the generated mutants (faulty programs) and original program are executing with the corresponding test data. One of the main drawbacks of mutation testing is its computational cost. Indeed, high execution time of mutation testing is a challenging research problem. Reducing the time and cost of mutation test is the main objective of this paper. In the traditional mutation methods and tools the mutants are injected randomly in each instructions of a program. Meanwhile, in the real-world program, the probability of fault occurrences in the simple locations (instructions and data) of a program is negligible. With respect to the 80-20 rule, 80% of the faults are found in 20% of the fault-prone code of a program. In the first stage of the proposed method, Artificial Bee Colony optimization algorithm is used to identifying the most fault prone paths of a program; in the next stage, the mutation operators (faults) are injected only on the identified fault-prone instructions and data. Regarding the results of conducted experiments on the standard benchmark programs, Compared to existing methods, the proposed method reduces 28.10% of the generated mutants. Reducing the number of generated mutants will reduce the cost of mutation testing. The traditional mutation testing tools (Mujava, Muclipse, Jester, Jumble) can perform the mutation testing with a lower cost using the method presented in this study.