Speeding Up Model Checking of Timed-Models by Combining Scenario Specialization and Live Component Analysis ⋆ V´ ıctor Braberman 1 , Diego Garbervestky 1 , Nicol´as Kicillof 2 , Daniel Monteverde 1,3 , and Alfredo Olivero 3,4 1 DC, FCEyN, UBA - Argentina. {vbraber,diegog}@dc.uba.ar 2 Microsoft - USA. nicok@microsoft.com 3 INTEC, UADE - Argentina. {damonteverde,aolivero}@uade.edu.ar 4 ECyT, UNSAM - Argentina. Abstract. The common practice for verifying properties described as event occurrence patterns is to translate them into observer state ma- chines. The resulting observer is then composed with (the components of) the system under analysis in order to verify a reachability property. Live Component Analysis is a “cone of influence” abstraction technique aim- ing at mitigating state explosion by detecting, at each observer location, which components are actually relevant for model checking purposes. In- terestingly enough, the more locations the observer has, the more precise the relevance analysis becomes. This work proposes the formal under- pinnings of a method to safely leverage this fact when properties are stated as event patterns (scenarios). That is, we present a sound and complete method of property manipulation based on specializing and complementing scenarios. The application of this method is illustrated on two case studies of distributed real-time system designs, showing dra- matic improvements in the verification phase, even in situations where verification of the original scenario was unfeasible. 1 Introduction The use of observers to express properties for the automatic verification of mod- els of reactive software is common-place (e.g., [2, 1, 12], etc). This is specially the case when requirements are heavily based on event occurrences, since using logical formalisms tends to be cumbersome, if possible at all [12]. In most cases, (automaton) observers are either hand-written or generated from some high-level property specification notation (for example, event patterns). Once defined, ob- servers are composed with the components of the System Under Analysis (SUA), and a reachability property is verified using a model checker. Live Component Analysis (LCA) is a special case of cone-of-influence abstrac- tion [13] that works by instructing a model checker to ignore the state of some ⋆ Research partially supported by CONICET and the projects UBACyT X021, ANPCyT PICTO-CRUP 31352 and ANPCyT PICT 32440