An Experimental Comparison of Safety-Specific Versus Non-Safety-Specific Software Methods

Abstract

Due to the rapid increase in computer-related technologies, various organizations are continuing to put higher demands on software systems. These demands are sometimes placing software in total or partial control over critical system functions such as navigating planes, controlling traffic systems, identifying military targets, and facilitating medical treatments to name just a few. Unfortunately, certain faults in such a system can result in catastrophic consequences such as death, injury, or environmental harm. To help ensure that software systems are safe, developers use safety-specific techniques in addition to more general non-safety-specific techniques. However, no experimental studies exist that compare safety-specific versus non-safety-specific techniques. Such experiments are necessary in order to determine whether or not developers using safety-specific software techniques are able to produce software systems with statistically fewer latent hazardous faults than those developers not using these techniques. Furthermore, such experiments are necessary in order to determine weaknesses in current methods, thus pinpointing areas for future research. In order to provide an initial data point in this area, the authors conducted an experiment that compared certain safety-specific and non-safety-techniques at developing a small safety-critical software system. This paper discusses this experiment, presents its results, and statistically analyzes the results.