A foundation for runtime monitoring

Abstract

Runtime Verification is a lightweight technique that complements other verification methods in an effort to ensure software correctness. The technique poses novel questions to software engineers: it is not easy to identify which specifications are amenable to runtime monitor-ing, nor is it clear which monitors effect the required runtime analysis correctly. This exposition targets a foundational understanding of these questions. Particularly, it considers an expressive specification logic (a syntactic variant of the modal μ-calculus) that is agnostic of the verification method used, together with an elemental framework providing an operational semantics for the runtime analysis performed by monitors. The correspondence between the property satisfactions in the logic on the one hand, and the verdicts reached by the monitors performing the analysis on the other, is a central theme of the study. Such a correspondence underpins the concept of monitorability, used to identify the subsets of the logic that can be adequately monitored for by RV. Another theme of the study is that of understanding what should be expected of a monitor in order for the verification process to be correct. We show how the monitor framework considered can constitute a basis whereby various notions of monitor correctness may be defined and investigated.