Discrete Bayesian-Based Scheme of Software Reliability Demonstration Testing with Decreasing Function for the Safety Critical Software

Abstract

The Bayesian-based scheme of software reliability demonstration testing (SRDT) is suitable for the safety-critical software (SCS) for it can utilize the prior information to significantly decreasing the required testing effort of SRDT. Selecting the appropriate prior distribution of the failure probability is very significant to the discrete Bayesian-based scheme of SRDT. Utilizing the decreasing function for constructing the prior distribution is a new and effective approach for Bayesian theory. Thus this paper uses the decreasing function of the failure probability as its prior distribution and then proposes a new Bayesian-based scheme with prior information based on decreasing function (DFBSRDT). Using three failure data-sets as the prior information, we compare the proposed DFBSRDT with two existing Bayesian-based schemes, i.e. the Bayesian-based schemes without prior information (BSRDT) and with prior information (PIBSRDT). The results show that the proposed DFBSRDT is more effective and applicable than BSRDT and PIBSRDT for significantly decreasing the required testing effort of SRDT for SCS especially the SCS with high reliability.