We present a framework for reasoning about abstract open systems. Open systems, also called "reactive systems" or "modules", are systems that interact with their environment and whose behaviors depend on these interactions. Embedded software is a typical example of open system. Module checking [KV96] is a verification technique for checking whether an open system satisfies a temporal property no matter what its environment does. Module checking makes it possible to check adversarial properties of the "game" played by the open system with its environment (such as "is there a winning strategy for a malicious agent trying to intrude a secure system?"). We study how module checking can be extended to reason about 3-valued abstractions of open systems in such a way that both proofs and counter-examples obtained by verifying arbitrary properties on such abstractions are guaranteed to be sound, i.e., to carry over to the concrete system. We also introduce a new verification technique, called generalized module checking, that can improve the precision of module checking. The modeling framework and verification techniques developed in this paper can be used to represent and reason about abstractions automatically generated from a static analysis of an open program using abstraction techniques such as predicate abstraction. This application is illustrated with an example of open program and property that cannot be verified by current abstraction-based verification tools.
展开▼