Integrating Safety Analysis with Automatic Test-Data Generation for Software Safety Verification

摘要

Typically verification focuses on demonstrating consistency between an implementation and a functional specification. For safety critical systems this is not sufficient, the implementation must also meet the system safety constraints and safety requirements. The work presented in this paper builds on the authors' previous work in developing a general framework for dynamically generating test-data using heuristic global optimisation techniques. This framework has been adapted to allow automated test-data generation to be used to support the application of software fault tree analysis. Using the framework a search for test-data that causes an identified software hazard condition can be performed automatically. The fact that a hazardous condition can arise may be discovered much earlier than with conventional testing using this automated approach. If no test-data is located then SFTA, or other forms of static analysis, can be performed to give the necessary assurance that no such data exists. A number of extensions to this basic approach are also outlined. These are, integration with fault injection, testing for exception conditions and testing for safe component reuse and integration. Preliminary results are encouraging and show that the approach justifies further research.