Validation through exclusion: techniques for ensuring software safety

An approach to software safety is presented that emphasizes the construction of a test set for checking that conditions leading to software safety failures do not occur. The approach combines some mutation analysis along with some recent work on the fundamentals of testing. This approach tests over a very constrained space of functions where the test sets are determined by a type of software fault tree analysis. It is shown that test sets can be derived with the guarantee that if the system passes its tests, then it is correct (i.e., safe) or suffers from some safety flaw not reflected in the model.<
>