Modeling the Reliability of Existing Software using Static Analysis

Software unreliability represents an increasing risk to overall system reliability. As systems become larger and more complex, mission critical and safety critical systems have had increasing functionality controlled exclusively through software. This change, coupled with generally increasing reliability in hardware modules, has resulted in a shift of the root cause of systems failure from hardware to software. Market forces, including decreased time to market, reduced development team sizes, and other factors, have encouraged projects to reuse existing software as well as to purchase COTS software solutions. This has made the usage of the more than 200 existing software reliability models increasingly difficult. Traditional software reliability models require significant testing data to be collected during software development in order to estimate software reliability. If this data is not collected in a disciplined manner or is not made available to software engineers, these modeling techniques can not be applied. It is imperative that practical reliability modeling techniques be developed to address these issues. It is on this premise that an appropriate software reliability model combining static analysis of existing source code modules, limited testing with path capture, and Bayesian belief networks is presented. Static analysis is used to detect faults within the source code which may lead to failure. Code coverage is used to determine which paths within the source code are executed as well as how often they execute. Finally, Bayesian belief network is then used to combine these parameters and estimate the resulting software reliability