Reducing Verification Costs through Practical Formal Methods: A Survey

Verification of software can be as much as 88% of the total cost to deliver a high dependability system. Significant manual effort is often required to produce required verification evidence. Formal methods are believed to hold promise by providing a more automatic means of verification. Advances have been made in theorem provers and model checkers aimed to support the verification efforts, however there are several challenges in producing verification evidence when the requirement and design specifications use nonlinear and floating point constraints mixed with linear, logical, and bit constraints. Some tools apply strategies to produce tests to support verification, but the fault-finding effectiveness of the generated tests is questionable. Finally, practical means for composing formal specifications is important for constructing specifications that scale to large systems. This paper discusses the challenges, summarizes needed capabilities of formal method technologies, provides a survey of tools, with experiments to assess and compare capabilities, and finally discusses future needs to address some of the challenges. Keyword: model-based testing, formal methods, theorem proving, test generation, constraint solving, model checking