Software hazard mining

Abstract

Software fault injection is a process that discovers how "badly" software can behave after its state gets corrupted. Fault injection results are sometimes viewed suspiciously since state corruption is hypothetically-based. In response to this suspicion, this paper explores the potential return-on-investment when artificial state corruptions are used. We will primarily focus on fault injection's unique ability to reveal hazards that were inadvertently overlooked during software requirements and design. To our knowledge, this application of fault injection has never been exploited.