Kiril Bogdanov

Title: State machine inference in support of testing

Abstract: Testing based on finite-state machines (FSM) can be used as a
proof of conformance of an implementation under test to a model. Such a
model is traditionally expected to be derived from a specification,
which in many practical applications is out of date or completely
missing. Inference methods make it possible to take traces from an
existing system and build a hypothesis model. One can subsequently
analyse the inferred model and generate tests to cover what has not been
observed in the traces it was built from. In a way, inference
complements testing so that if a test suite is good, even a mediocre
inference method should construct a correct model from the tests. The
real challenge for an inference method is to build a good model even
where relatively few traces are available. The talk will discuss test
generation, the problem of model inference and the work I do in
construction of models.