Building statistical test-cases for smart device software - an example

Statistical testing (ST) of software or logic-based components can produce dependability information on such components by yielding an estimate for their probability of failure on demand. An example of software-based components that are increasingly used within safety-related systems e.g. in the nuclear industry, are smart devices. Smart devices are devices with intelligence, capable of more than merely representing correctly a sensed quantity but of functionality such as processing data, self-diagnosis and possibly exchange of data with other devices. Examples are smart transmitters or smart sensors. If such devices are used in a safety-related context, it is crucial to assess whether they fulfil the dependability requirements posed on them to ensure they are dependable enough to be used within the specific safety-related context. This involves making a case for the probability of systematic failure of the smart device. This failure probability is related to faults present in the logic or software-based part of the device. In this paper we look at a technique that can be used to establish a probability of failure for the software part of a smart monitoring unit. This technique is "statistical testing" (ST). Our aim is to share our own experience with ST and to describe some of the issues we have encountered so far on the way to perform ST on this device software.