Undetected faults in protocol testing

We investigate ways in which UIO-based conformance testing can fail to catch faults, including single and multiple faults, faults with extra or missing states, and faults at both the test sequence and subsequence levels. Given a particular error and test method, the error is masked if it is not detected by the test method. Many forms of fault masking are possible, and all test methods we have considered exhibit some forms of masking. Faults captured at the test subsequence level may become masked at the sequence level, and vice versa. Fault masking has been used to argue relative merits of various testing methods. Because of the pervasiveness of masking, we cannot use masking alone to argue that one UIO-based test method is superior to another. Information about the density of masked faults among all faults is needed to evaluate a test method.<
>