The effect of correlated faults on software reliability

The reliability models often assume random testing and statistical independence of faults to keep the analysis tractable. In practice, these assumptions do not hold. This paper presents a reliability modeling approach that considers nonrandom testing. This approach is used to calculate the fault exposure ratio, which characterizes the testing process. The analysis of the experimental data suggests that the fault exposure ratio varies differently in the early and the later stages of testing. The analysis here presents an explanation of this behavior.