Statistical approach to software reliability certification

We present a sequential software release procedure that certifies with some confidence level that the next error is not occurring within a certain time interval. Our procedure is defined in such a way that the release time is optimal for single stages and the global risk can be controlled. We assume that the failure detection process can be modeled as a General Order Statistics model where its unknown parameters are studied from a Bayesian point of view. We show how to implement our procedure and study its performance via simulation