Abstract

Deductive program verifiers attempt to construct a proof that a given program satisfies a given specification. Their implementations reflect the semantics of the programming language and the specification language, and often include elaborate proof search strategies to automate verification. Each of these components is intricate, which makes building a verifier from scratch complex and costly.

In this lecture series, we will present an approach to build program verifiers as a sequence of translations from the source language and specification via intermediate languages down to a logic for which automatic solvers exist. This architecture reduces the overall complexity by dividing the verification process into simpler, well-defined tasks, and enables the reuse of essential elements of a program verifier such as parts of the proof search, specification inference, and counterexample generation. We will introduce intermediate verification languages and demonstrate how they can encode interesting verification problems.