An empirical study on reliability modeling for diverse software systems

Reliability and fault correlation are two main concerns for design diversity, yet empirical data are limited in investigating these two. In previous work, we conducted a software project with real-world application for investigation on software testing and fault tolerance for design diversity. Mutants were generated by injecting one single real fault recorded in the software development phase to the final versions. In this paper, we perform more analysis and experiments on these mutants to evaluate and investigate the reliability features in diverse software systems. We apply our project data on two different reliability models and estimate the reliability bounds for evaluation purpose. We also parameterize fault correlations to predict the reliability of various combinations of versions, and compare three different fault-tolerant software architectures.