Approximate Model Checking




  • Probabilistic abstraction .
    Given a program and a property to check, which admits some tester, we study how to find, in a general way, some probabilistic abstraction, defined in [LLMPR]. Such an abstraction is a program transformation allowing to abstract certain variables with respect to the tester. We can then apply the test to concrete programs and approximatively check a property in constant time.

  • Probabilistic Model Checking.
    Model Checking has been recently extended to probabilistic protocols, which can be represented by probabilistic transition systems (Markov chains). The objective is to compute the satisfaction probability of some temporal property as reachability, liveness or safety. The main problem is the space complexity of these methods. A natural issue is to approximate this probability. Although the problem is not approximable in full generality, there exist some randomized approximation schemes in polynomial time for certain classes of properties [LP]. We apply some sampling techniques, as Monte-Carlo method, to probabilistic protocols in order to generate randomized execution paths and approximatively verify these quantitative properties.

    Publications