Mutation Testing for Rule-Based Verification of Railway Signaling Data

Abstract

Industry applications of formal verification to signalling control tables require formulation of a large number of mathematical conjectures expressing verification rules. It is paramount to establish the validity and completeness of these conjectures. The paper discusses a mutation based validation technique that guides domain experts in the construction of such verification rules. Furthermore, we use genetic programming to quickly generate millions of well-formed data mutations of control tables and to synthesise mutation programs. The technique is illustrated by a synthetic running example and a discussion of our experience in using it in the industrial setting.