Trap-Based Software Review: How Will Engineers Embed Bugs Due to Cognitive Error?

摘要

Reviewing is a key technology to software reliability engineering. One of the most important purposes of reviewing is to detect faults. There are various sorts of reviewing techniques proposed such as Defect-based Reading and Orthogonal Defect Classification, which focus on fault. In this research, we focus on "trap" in a developer's cognitive process. A trap is a part of a deliverable or its pattern that induces an engineer to make an error and to embed a fault. Firstly, this paper models the cognitive process called Software Trap model and proposes Trap-based Review (TBR). Secondly, we introduce three traps which were extracted from commercial software development. Then the steps for TBR are also explained. Finally, we show case studies to evaluate effectiveness of TBR in commercial software product in a financial domain. The result showed that TBR succeeded in detecting the faults that were not detected through commercial software development. This paper also mentions how the trap-based approach can be applied to dynamic testing.