A bayesian differential debugging model for software reliability

An assumption commonly made in early models of software reliability is that the failure rate of a program is a constant multiple of the number of bugs remaining. This implies that all bugs have the same effect upon the overall failure rate. The assumption is challenged and an alternative proposed. The suggested model results in earlier bug-fixes having a greater effect than later ones (the worst bug show themselves earlier and so are fixed earlier), and the DFR properly between bug-fixes (confidence in programs increases during periods of failure-free operation, as well as at bug-fixes). The model shows a high degree of mathematical tractability, and allows a range of reliability, and allows a range of reliability measures to be calculated exactly. Predictions of total execution time to achieve a target reliability, are obtained.