Abstract: In order to reduce the total number of the testing cases but not to decrease the confidence level of the testing results for the reliability demonstration of safety-critical software, a new method which uses Bayesian inference with prior knowledge dynamic integration is presented on the basis of analyzing the classical statistical hypothesis testing and the ignorance prior Bayesian method. In particular, the problem of how to determine the parameters of the prior distribution for the density function of software failure probability is solved. Test shows that it requires less testing cases to achieve the same confidence level by using the method provided in this paper.

Key words: software reliability, reliability demonstration, Bayesian method, safety-critical software, software testing