Detection of feature interactions in intelligent networks by verification

The feature interaction problem in Intelligent Networks (IN) increasingly obstructs the rapid introduction of new features. Checking each feature manually against every other feature is no longer feasible. We give an overview on current (verification) approaches for the off-line detection of feature interactions, and we categorize them into specified-property approaches (which verify two separate formal descriptions against each other) and general-property approaches (which require only one description). We improve the general-property approaches by presenting a formal framework that allows identification of all potential feature interactions in a specified system. It is centered around formal definitions of the notions of ''feature'' and ''feature interaction'' which cover the functional aspects in a system. The definition of feature interaction is based on the notion of ''behaviour of a feature''. The telephone system is modeled by a global, structured automaton. Adding a feature is realized by adding transitions and (possibly) extending the state space, employing a specific specification style. Based on the formal definition, we derive more sophisticated detection criteria. There is already some tool support. A first case study found not only the already known feature interactions but also two interferences which we overlooked during specification. A second, extended case study is still in progress. Since the general-property approaches complement the specified-property approaches in a certain sense, we propose to apply both approaches together in order to solve the feature interaction detection problem.