Static Analysis and Veri cation of Aerospace Software by Abstract Interpretation (Abstract)

The validation of software checks informally (e.g., by code reviews or tests) the conformance of the software executions to a speci cation. More rigorously, the veri cation of software proves formally the conformance of the software semantics (that is, the set of all possible executions in all possible environments) to a speci cation. It is of course di cult to design a sound semantics, to get a rigorous description of all execution environments, to derive an automatically exploitable speci cation from informal natural language requirements, and to completely automatize the formal conformance proof (which is undecidable). In model-based design, the software is often generated automatically from the model so that the certi cation of the software requires the validation or veri cation of the model plus that of the translation into an executable software (through compiler veri cation or translation validation). Moreover, the model is often considered to be the speci cation, so there is no speci cation of the speci cation, hence no other possible conformance check. These di culties show that fully automatic rigorous veri cation of complex software is very challenging and perfection is impossible. We present abstract interpretation and show how its principles can be successfully applied to cope with the above-mentioned di culties inherent to formal veri cation.