Significant Diagnostic Counterexamples in Probabilistic Model Checking

Abstract. This paper presents a novel technique for counterexample generation in probabilistic model checking of Markov chains and Markov Decision Processes. (Finite) paths in counterexamples are grouped together in witnesses that are likely to provide similar debugging information to the user. We list five properties that witnesses should satisfy in order to be useful as debugging aid: similarity, accuracy, originality, significance, and finiteness. Our witnesses contain paths that behave similarly outside strongly connected components. Then, we show how to compute these witnesses by reducing the problem of generating counterexamples for general properties over Markov Decision Processes, in several steps, to the easy problem of generating counterexamples for reachability properties over acyclic Markov chains.

• Type: Journal
• Year: 2008
• Where: CORR
• Authors: Miguel E. Andrés, Pedro R. D'Argenio, Peter van Rossum
• Keywords: CORR 2008, Counterexamples, Education, Markov Chains, Markov Decision