Inferring Sufficient Conditions with Backward Polyhedral Under-Approximations

In this article, we discuss the automatic inference of sufficient pre-conditions by abstract interpretation and sketch the construction of an under-approximating backward analysis. We focus on numeric domains and propose transfer functions, including a lower widening, for polyhedra, without resorting to disjunctive completion nor complementation, while soundly handling non-determinism. Applications include the derivation of sufficient conditions for a program to never step outside an envelope of safe states, or dually to force it to eventually fail. Our construction is preliminary and essentially untried, but we hope to convince that this avenue of research is worth considering.