2014

Tyrväinen, Tero

Littlewood and Wright presented a Bayesian belief network model for software reliability analysis in their article The use of multilegged arguments to increase confidence in safety claims for software-based systems: A study based on a BBN analysis of an idealized example. In the model, the confidence on the software's reliability depends on testing and verification results and the prior confidence on the software specification and the "oracle" used in testing. Littlewood and Wright introduced counterintuitive results: testing or verification can reduce the confidence on the software's reliability even if no faults are found. This document provides an explanation why the model produces these counterintuitive results. The results indicate that the counterintuitive results do not completely depend on the calculation formulas and are in theory possible with more comprehensive models as well.