Slicing Executable System-of-Systems Models for Efficient Statistical Verification

Abstract

A System of Systems (SoS), composed of independent constituent systems, can create synergy among its systems to achieve a common goal. Many studies have used statistical model checking techniques to verify how well an SoS can achieve its goals. SoS models are usually complex and probabilistic, which makes statistical verification computationally expensive. To reduce this cost, dynamic slicing techniques can be applied to SoS models since both dynamic slicing and statistical verification focus on the models' execution samples. However, existing dynamic slicing techniques cannot guarantee executable accurate slices of SoS models when the models contain uncertainty. Therefore, we propose a hybrid slicing approach that combines dynamic backward slicing and modified observation-based slicing to produce accurate executable slices. Experimentation on the proposed technique found that the verification time was significantly reduced (47--56%), depending on the property, while preserving the verification results.