GUIDING SEARCH-BASED SOFTWARE TESTING WITH DEFECT PREDICTION: AN EMPIRICAL INVESTIGATION

Search-Based Software Testing (SBST) has emerged as a powerful technique for automated test case generation, effectively achieving high code coverage. However, maximizing code coverage does not always correlate directly with the ability to detect real faults. This paper presents an empirical investigation into the effectiveness of using theoretical defect predictors to guide the search process in SBST, aiming to enhance its fault-finding capability. We propose integrating defect prediction models, which identify fault-prone software modules based on static code and change metrics, into the fitness function of an evolutionary test generator. Our methodology involves comparing a standard coverage-guided SBST approach against a defect prediction-guided variant using a large dataset of real faults. Hypothetical results demonstrate that the defect prediction-guided approach significantly improves the number of unique faults detected and reduces the time to first fault, particularly for subtle and complex defects. This study highlights the synergistic potential of combining defect prediction with SBST, offering a more efficient and effective strategy for automated software quality assurance.