A Bayesian estimation method for the failure rate of a possibly correct program

An extension of software reliability modeling is introduced to account for the possibility of programs which, after some debugging, contain no more errors. This is achieved by a repetitive application of the Bayes law, each time taking the posterior of the last step as a prior for the next one. A class of conjugate priors considerably facilitates this modeling. The resulting model includes an estimator for the probability that a program still contains errors, which is an upper bound for the failure probability.>