Abstract
With progression of the digital age, the complexity of software continues to grow. As a result, methods to quantitatively assess characteristics of software have attracted significant attention. These efforts have led to a large number of new measures such as coupling metrics, many of which seek to consider the impact of correlations between components and failures on application reliability. However, most of these approaches set the coupling parameters arbitrarily by making assumptions instead of utilizing experimental data and therefore may not accurately capture actual coupling between components of software application. Since the coupling matrix is often set arbitrarily, the existing approaches to assess software reliability considering component correlation fail to reflect the real degree of interaction and relationships among software components. This paper presents an efficient approach to assess the software reliability considering correlated component failures, incorporating software architecture while considering actual internal coupling of software with an efficient approach based on multivariate Bernoulli (MVB) distribution. The unified framework for software coupling measurement is informed by a comprehensive survey of frameworks for object-oriented and procedure-oriented software. This framework enables the extraction of more accurate coupling among components. The effectiveness of this method is illustrated through an experimental study by applying it to a real-time software application.