A Software Reliability Model Considering the Initial and Test-Start Faults with Dependent Failures

With the Fourth Industrial Revolution, software has evolved into complex large-scale systems. Because of the complexity of these systems, failures and defects in individual software components can propagate through the entire system, potentially causing substantial losses. Numerous studies have aimed to address this issue and improve software reliability. In this paper, we extend previous research by proposing NHPP-based software reliability models that consider dependent failures and initial faults. Furthermore, considering failures in both initial and test stages, we propose software reliability models suitable for real-world problems. The proposed models, incorporating these two aspects, were compared with 21 traditional software reliability models and three datasets using nine criteria. The results indicate that accounting for dependent failures and initial faults improves software reliability modelling in complex systems. The model considering initial failures at the start demonstrated the best performance on two datasets, whereas the model considering initial failures during the testing phase also showed strong results on one dataset. Furthermore, both proposed models demonstrated excellent performance across all three datasets. These results indicate that, in complex software systems, model performance depends on the treatment of initial faults and dependent failures.