Abstract Software faults are costly to find and remove from programs. It is better to avoid inserting the faults in the first place. The authors identify requirements‐related faults that can arise during the transformation of formal specifications into programs, analyse each to discover its causes, and propose corresponding fault prevention methods. The approach works by analysing a specification, decomposing it into specification fragments with different levels like components, determining the correct implementation order among the components, and enabling automated code generation for the components and other low‐level specification fragments. The approach seeks to reduce the faults in programs significantly by guiding programmers to work effectively and productively. The authors evaluate the approach's performance experimentally and compare it to the performance of a fault‐prevention method with industry applications.

Formal specifications can provide a solid foundation for software development and support for techniques of software quality assurance, such as specification-based inspection and testing. To ensure that these techniques can be applied effectively in practice, efficiently and accurately understanding specifications becomes extremely important. While this may be relatively easy for well-trained developers in formal methods, it can be rather difficult for computer since computer does not easily understand specifications. This difficulty poses a challenge for realizing automatic specification-based verification techniques that are in high demand for reducing development cost and improving software reliability. In this paper, we address this problem by discussing how the formal specification can be transformed into a knowledge graph to provide comprehensible, well-organized details of the specification for developers and computers. The transformation is done by extracting and storing information about attributes of each component and by establishing relationships between components in a formal specification. We elaborate on a top-down approach of constructing a knowledge graph from a specification, including creating an ontology, designing the Entity–Relationship (ER) diagram of the relational database based on the created ontology, extracting and storing attribute and relationship information in the relational database, mapping ontology to its instances and relational data to RDF triples, and displaying knowledge graph. Further, we present a case study to show how our approach works on the formal specification of an ATM system. Finally, we describe three experiments to evaluate its performance in improving specification readability, effectively guiding inspectors to establish traceability links between specifications and programs, and detecting defects through program inspection, respectively.

Linking the components in a formal specification to those in the corresponding program is a prerequisite for formal specification-based program fault detection. Existing traceability link techniques for reducing manpower and time cost suffer from the limitation in effectiveness due to over dependency of textual similarity. Unlike the existing work, this paper presents an automatic method for constructing traceability links between SOFL formal specifications and Java codes, taking semantical, structural, functional, and relational similarities measures into account. It operates at multiple levels of a formal specification, such as data flows, processes, and modules, to establish finegrained link relationships between artifacts. Further, a comparative evaluation of the proposed method, using two selected modules of the SOFL formal specification of a critical ATM system and its Java implementation with 951 code of lines, demonstrates an improvement in precision and more generality than existing latent semantic indexing that is an information retrieval-based method.