The Key Techniques Research Of EFSM-based Automated Test Case Generation

The rapid development of information technologies makes the scale of the soft-ware becoming even bigger and the version upgrading faster and faster, the possibility of introducing faults is also growing. In order to minimize the faults in the software as far as possible, researchers and developers have proposed lots of software quality assurance means. However, software testing, which plays an irreplaceable role and attracts more and more attention, is still a most important and effective way to ensure software quality and reliability.Software testing is known to be expensive and time-consuming, hence automated testing has become a tendency due to its capability of reducing testing cost and increas-ing software quality. Moreover, the key issue of software testing is automated test case generation. One of the common approaches of automated test case generation is to create a model and utilize the model to generate test cases. The Extended Finite State Machine (EFSM) model attracts increasing attention in terms of its strong express a-bility and easy understandable feature. In EFSM-based testing, the automated test data generation, which closely related to the path feasibility problem, is still a challenge. Although there are some existing works, EFSM-based test data generation techniques are still far from mature. In addition, the present studies less consider the test suite optimization and the automated test oracle generation problem.To address the aforementioned problems, this thesis proposes a novel approach named ATGEM. In order to address infeasible issue, we propose a metric to predict the infeasible probability so as to bypass the infeasible paths as far as possible and im-prove the test case generation efficiency. Afterwards, a fitness function is designed by collecting the run-time feedback information based on building an executable model. Hence the scatter search algorithm is introduced to pick the path from the sorted can-didate path set to generate test cases automatically. Through the utilization of semantic analysis technique, the executable model can be executed like a program, and it still remains the ability of abstract expression. By means of run-time feedback information based on executable model, data type differences problem can be ignored in fitness function design. Therefore, this fitness function can be applied to many data types and has a wide range of applications. Another advantage of the executable model is that the corresponding outputs associated with generated test data are also collected to generate oracle information automatically. Moreover, the ATGEM attempts to find a test suite that has fewer paths, longer path length and goodness feasibility by means of above sub-methods iteration to meet the specified coverage criterion, since the previ-ous study shown that the test suite with a small number of longer test cases improves fault-detection efficiency compared to shorter ones.The path feasibility metric of ATGEM is designed carefully to trade-off path length and feasible evaluation value, since a longer path has a higher probability of infeasible due to the fact that a longer path may contain more conflict conditions a-mong transitions. However, the metric is still not precise enough. Therefore, we propose an approach (named ATGEMmop) based on multi-objective optimization tech-nique to solve the path ordering problem. Two fitness functions are designed to obtain the Pareto-optimal solutions.We design a series of experiments with the goal of verifying ATGEM and ATG-EMmop. Two approaches apply to several EFSM models in order to verify the effective-ness and efficiency of test case generation, path feasibility metric and multi-objective optimization. The experiments also detect the number of infeasible paths in the model and analyze the possible reasons. The experimental results show that ATGEM have good effectiveness in path feasibility metric, generating optimized test suite and gen-erating test case to meet the specified coverage criterion. In addition, ATGEM is more efficient than the existing method in test data generation. Finally, the experiment re-sults show that ATGEMmop is superior to ATGEM except slightly lower time efficiency and has a better performance in finding a test suite that has fewer but longer test cases.