A Bayesian Belief Network Model for Early Prediction of Reliability for Computer-Based Safety-Critical Systems

Abstract

Computer-based safety-critical systems (CBSCS), found in automotive, nuclear power plants (NPP), space, health-care, etc. rely mainly on functional requirements and timing correctness. These systems are highly reactive and concurrent and demand not only safe and reliable systems but systems that remain secure and availabel while under attacks. Researchers and academicians have proposed various software reliability growth models (SRGM) and probabilistic models to quantify reliability and other dependability attributes. However, SRGM and the probabilistic models' accuracy depend on the sufficiency of failure data. The SCS development model follows rigorous design and development steps. Therefore, a very less number of failures occur during the testing or operational phase. Due to the non-sufficiency of failure data, the existing reliability growth models or the probabilistic models fail to accurately estimate or predict the reliability accurately. This paper presents a novel approach towards predicting the reliability of an SCS using the Bayesian Belief Network Model (BBN). The current approach takes the quality attributes of each and every phase of the Software Development Life Cycle (SDLC) model and hence gives a more accurate estimation.