A Functional Framework for Result Checking

Result checking is a general methodology for ensuring that untrusted computations are valid. Its essence lies in defining efficient checking procedures to verify that a result satisfies some expected property. Result checking often relies on certificates to make the verification process efficient, and thus involves two strongly connected tasks: the generation of certificates and the implementation of a checking procedure. Several ad-hoc solutions exist, but they differ significantly on the kind of properties involved and thus on the validation procedure. The lack of common methodologies has been an obstacle to the applicability of result checking to a more comprehensive set of algorithms. We propose the first framework for building result checking infrastructures for a large class of properties, and illustrate its generality through several examples. The framework has been implemented in Haskell.