A point-process stochastic model for software reliability

Abstract A formal approach for evaluating the reliability of computer software is through probabilistic models and statistical methods. This paper presents the former direction and is based on the ideas of the input-domain method. The software reliability function is introduced in a general way and is defined by a time-dependent operational profile and distributions of bugs in the input space. The assumption of uniformity leads to the Jelinski–Moranda model of software reliability. Another assumption of input space being organized as one path leads to a modified renewal model for the debugging process description. This model shows the natural increasing of time periods between failures. Asymptotic results on the number of failures are presented in conformity with some conventional empirical models of software reliability.