Incorporating domain-specific insights for automated test generation

Abstract

There has been growing discussion in Automated Test Generation (ATG) among both academia and industry due to its critical role in software development lifecycle aimed at enhancing the efficiency and quality of software testing. With the increased complexity in software systems and incessant demand for faster development cycles, more and more studies have been dedicated to purposing and improving existing techniques. Central to our investigation is the recognition of existing ATG methodologies’ limitations, particularly their effectiveness in real-world fault detection. Through an extensive review of literature and existing techniques, we identify a critical gap in the application of domain knowledge - specific insights derived from manual bug-fixing processes that are not yet leveraged in existing ATG approaches. In this thesis, we introduce an approach to integrate this untapped domain knowledge with automated testing processes. Our approach aims to complement existing ATG methodologies by harnessing developers’ experimental insights during manual debugging for test generation. The test cases generated by our approach are destined to be different from those generated by existing methodologies due to the difference in test input collection. We evaluate our technique using subjects from Defects4J. Our evaluation demonstrates the potential of this integrated approach, showing an improvement in fault detection capabilities and test coverage. We also discuss the broader implications of our findings for software development practices, emphasizing the value of synergizing human expertise with automation for advancing software quality assurance. This work contributes to the field of ATG by offering a new perspective on improving automated testing’s effectiveness and efficiency. It opens avenues for future research on the integration of human-derived insights with automated systems, marking a step forward in the evolution of software testing methodologies.

Date of Award	2024
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Shing Chi CHEUNG (Supervisor)