Model-based statistical testing (MBST) is a black-box testing technique that evaluates components or systems with the use of statistics. MBST can be seen as a controlled statistical experiment. During this experiment, the relevant and representative test cases are chosen from a large number of possible tests. One focus of the method is the construction of a statistical test model which describes the functional system requirements as a set of input-output-trajectories. The test model is constructed from the original requirements by the sequence-based specification (SBS), a systematic inspection method. This systematic inspection allows finding errors, inconsistencies, missing requirements, and contradictions in the original requirements. If the test object or its requirements are changed, the only thing that needs to be done is to adapt one artifact, i.e., the test model, instead of changing many test cases.

The input-output-sequences of the test model are annotated with their frequency of occurrence or the criticality, the expected system outputs, and executable scripts (needed for automated test execution). A test case is a path through the test model. Test cases are automatically generated from the test model by different strategies including model coverage tests and random tests. Model coverage tests make sure that the whole model is covered by test cases in the minimum number of test cases and test steps. Typical coverage criteria are: state coverage and transition coverage. Randomly generated tests represent realistic test cases according to the selected profile. The results from the random tests can be used to estimate the reliability of the current status of the test object.

The reliability model used here estimates the probability that no failure will occur in future operation. The objective is to achieve reliability predictions close to 100%. Input parameters for the reliability estimations are the number of executions and the number of failures for every input. Reliability estimates are provided for single system inputs and input sequences.

http://tai.iese.fraunhofer.de