Model-based quality assurance of instrumented context-free systems

Abstract

The ever-growing complexity of today’s software and hardware systems makes quality assurance (QA) a challenging task. Abstraction is a key technique for dealing with this complexity because it allows one to skip non-essential properties of a system and focus on the important ones. Crucial for the success of this approach is the availability of adequate abstraction models that strike a fine balance between simplicity and expressiveness. This thesis presents the formalisms of systems of procedural automata (SPAs), systems of behavioral automata (SBAs), and systems of procedural Mealy machines (SPMMs). The three model types describe systems which consist of multiple procedures that can mutually call each other, including recursion. While the individual procedures are described by regular automata and therefore are easy to understand, the aggregation of procedures towards systems captures the semantics of context-free systems, offering the expressiveness necessary for representing procedural systems. A central concept of the proposed model types is an instrumentation that exposes the internal structure of systems by making calls to and returns from procedures observable. This instrumentation allows for a notion of rigorous (de-) composition which enables a translation between local (procedural) views and global (holistic) views on a system. On the basis of this translation, this thesis presents algorithms for the verification, testing, and learning of (instrumented) context-free systems, covering a broad spectrum of practical QA tasks. Starting with SPAs as a “base” formalism for context-free systems, the flexibility of this concept is shown by including features such as prefix-closure (SBAs) and dialog-based transductions (SPMMs). In a comparison with related formalisms, this thesis shows that the simplicity of the proposed model types not only increases the understandability of models but can also improve the performance of QA tasks. This makes SPAs, SBAs, and SPMMs a powerful tool for tackling the practical challenges of assuring the quality of today’s software and hardware systems.