Software reliability estimation based on static error detection

The estimation of a programs' reliability is an essential part in the process of software development. Existing methods for the analysis of software reliability are based on run-time data, program metrics, and properties of development process or program architecture. The disadvantage of these methods is that they use indirect information about the errors, which are the main cause of program unreliability. In the paper we present a novel approach for software reliability estimation. This approach is based on error detection using static source code analysis. We extend static analysis with developed algorithms which calculate error probabilities and program reliability characteristics. The characteristics are the probability of successfull program termination, the probability of the program is operable after execution of n statements, and mean number of executed statements before failure. The suggested approach has been implemented in the AEGIS tool and tested in numerous real-world software projects.