Standalone Event-B Models Analysis Relying on the EB4EB Meta-theory

Abstract

Event-B is a state-based correct-by-construction system design formal method relying on proof and refinement where system models are expressed using set theory and First Order Logic (FOL). Through the generation and discharging of proof obligations (POs), Event-B natively supports the establishment of properties such as safety invariant, convergence and refinement. Other properties, relevant to system verification, may be studied as well, but need to be explicitly formalised by the designer, or expressed in another formal method. This process compromises reusability and is error-prone, especially on larger systems. Recently, the reflexive EB4EB framework has been proposed for formalising Event-B concepts as first-class objects. It allows manipulating these concepts using FOL and set theory in Event-B. In this paper, we propose a rigorous methodology for extending the EB4EB framework, to support new system analysis mechanisms associated to properties that are not natively present in core Event-B. Thanks to the reflexive nature of this framework, new generic and reusable system properties and their associated POs are expressed once and for all, and for any refinement level. For specific systems, designers instantiate these properties and the associated POs are automatically generated and submitted to Event-B's provers. This methodology is used to define three analyses: deadlockfreeness, invariant weakness analysis and reachability, all of which are demonstrated on a case study.