How can we estimate software reliability with a continuous-state software reliability model?

During the last three decades the stochastic counting (discrete-state) process models like non-homogeneous Poisson processes have described the software reliability growth phenomenon observed in the testing phase, and gained the popularity to explain the software debugging process. On the other hand, the continuous-state process models based on the Brownian motion processes have an advantage in terms of the goodness-of-fit test based on the information criteria, like AIC and BIC. The most critical point for the continuous-state process models is that the software reliability can not be well defined in their modeling framework. The purpose of this paper is to answer the titled question, that is, we propose two methods to define quantitatively the software reliability and the MTBSF (mean time between software faults) for a continuous-state software reliability model.