Software System Reliability Analysis

System reliability is a fundamental requirement for safety-critical weapons systems. A key challenge is identifying reliability problems early so that they can be fixed quickly and cheaply. Reliability problems are often integration problems: integration often reveals that flaws that seemed minor in isolated components but lead to serious system-wide reliability problems. The combination of static source code analysis with data mining offers the potential to identify reliability problems before integration. By applying data mining to the data generated through static analysis, we can recognize patterns that indicate when a software system is becoming a reliability risk. We propose a tool that will flag components that are likely to cause integration problems later in development, giving engineers time to investigate and solve problems before the expensive integration step. The proposed tool will collect metrics from the target system; learn patterns for correct and faulty system both passively and interactively; refine the patterns when new data is gathered; and issue reports pointing out software modules that pose reliability risks.