Perspectives on search strategies in automated test input generation

摘要

This paper presents an analysis framework to characterize an automated test input generation technique as three major components: neighborhood definition, neighborhood selection strategy, and bootstrap test inputs. The framework provides researchers a unified perspective to investigate and understand the characteristics, strengths, and weaknesses of existing test input generation techniques: fuzzing, DSE, and their hybrids. Supported with empirical study results, the paper calls for future efforts towards more effective test input generation techniques.