Automatic String Test Data Generation For EFSM Model

With the expansion of software application area and the scale of software, if the quality of software cannot be guaranteed, it may cause huge loss of both life and property. Software testing is an important technology for ensuring reliability and security of the software, and automatic generation of testing data method is one of the most important testing technologies. But most methods of testing data generation only consider integer type and Boolean type, while seldom consider the string type.Automatic generation of testing data in string type is a difficult point and current research doesn’t have a good approach to solve this problem. The paper proposes a method of automatic generation of testing data in string type combining symbolic execution with constraint solving and genetic algorithm based on EFSM model. First, characters in a testing data’s candidate character set are encoded into an integer, and string constraints on the target path are converted to integer constraints, then solve the constraints and obtain an intermediate solution. Finally, testing data in string type is generated by genetic algorithm. The fitness function in genetic algorithm is used for variable in string type and related to the string’s distance. The comparison methods of string’s distance include hamming distance, edit distance and character distance, but all of them have some limits. This paper analyses these three methods’limits and then proposes a new method of comparison of string’s distance. This method can overcome the shortcomings of existing methods.Experimental results show that the proposed automatic generation of test data in string type method is more effective than genetic algorithm and simulated annealing on running time. With the increment string’s length, running time decreases exponentially. On the time of string testing data generation, new string comparison method proposed in this paper is also reduced more than edit distance and character distance.