Abstract Dependences for Alarm Diagnosis

Dependences for Alarm Diagnosis Xavier Rival ´Ecole Normale Sup´erieure 45, rue d’Ulm, 75230, Paris cedex 5, France We propose a framework for dependence analyses, adapted –among others– to the understanding of static analyzers outputs. Static analyzers like Astr´ee are sound but not complete; hence, they may yield false alarms, that is report not being able to prove part of the properties of interest. Helping the user in the alarm inspection task is a major challenge for current static analyzers. Semantic slicing, i.e. the computation of precise abstract invariants for a set of erroneous traces, provides a useful characterization of a possible error context. We propose to enhance slicing with information about abstract dependences. Abstract dependences should be more informative than mere dependences: first, we propose to restrict to the dependences that can be observed in a slice; we define dependences among abstract properties, so as to isolate abnormal behaviors as sourc...

• Type: Conference
• Year: 2005
• Where: APLAS
• Authors: Xavier Rival
• Keywords: Abstract Dependences, APLAS 2005, Static Analyzers, Static Analyzers Outputs