Deductive Runtime Certification
Konstantine Arkoudas, Martin Rinard · Electronic Notes in Theoretical Computer Science · 2005
This paper introduces a notion of certified computation whereby an algorithm not only produces a result r for a given input x, but also proves that r is a correct result for x. This can greatly enhance the credibility of the result: if we trust the axioms and inference rules that are used in the proof, then we can be assured that r is correct. Typically, the reasoning used in a certified computation is much simpler than the computation itself. We present and analyze two examples of certifying algorithms. We have developed denotational proof languages (DPLs) as a uniform platform for certified computation. DPLs integrate computation and deduction seamlessly, offer strong soundness guarantees, and provide versatile mechanisms for constructing proofs and proof-search methods. We have used DPLs to implement numerous well-known algorithms as certifiers, ranging from sorting algorithms to compiler optimizations, the Hindley-Milner W algorithm, Prolog engines, and more.