Abstract

Software development and evolution are highly distributed processes that involve a multitude of supporting tools and resources. Application programming interfaces are commonly used by software developers to reduce development cost and complexity by reusing code developed by third-parties or published by the open source community. However, these application programming interfaces have also introduced new challenges to the Software Engineering community (e.g., software vulnerabilities, API incompatibilities, and software license violations) that not only extend beyond the traditional boundaries of individual projects but also involve different software artifacts. As a result, there is the need for a technology-independent representation of software dependency semantics and the ability to seamlessly integrate this representation with knowledge from other software artifacts.

The Semantic Web and its supporting technology stack have been widely promoted to model, integrate, and support interoperability among heterogeneous data sources. This dissertation takes advantage of the Semantic Web and its enabling technology stack for knowledge modeling and integration. The thesis introduces five major contributions: (1) We present a formal Software Build System Ontology – SBSON, which captures concepts and properties for software build and dependency management systems. This formal knowledge representation allows us to take advantage of Semantic Web inference services forming the basis for a more flexibility API dependency analysis compared to traditional proprietary analysis approaches. (2) We conducted a user survey which involved 53 open source developers to allow us to gain insights on how actual developers manage API breaking changes. (3) We introduced a novel approach which integrates our SBSON model with knowledge about source code usage and changes within the Maven ecosystem to support API consumers and producers in managing (assessing and minimizing) the impacts of breaking changes. (4) A Security Vulnerability Analysis Framework (SV-AF) is introduced, which integrates builds system, source code, versioning system, and vulnerability ontologies to trace and assess the impact of security vulnerabilities across project boundaries. (5) Finally, we introduce an Ontological Trustworthiness Assessment Model (OntTAM). OntTAM is an integration of our build, source code, vulnerability and license ontologies which supports a holistic analysis and assessment of quality attributes related to the trustworthiness of libraries and APIs in open source systems.

Several case studies are presented to illustrate the applicability and flexibility of our modelling approach, demonstrating that our knowledge modeling approach can seamlessly integrate and reuse knowledge extracted from existing build and dependency management systems with other existing heterogeneous data sources found in the software engineering domain. As part of our case studies, we also demonstrate how this unified knowledge model can enable new types of project dependency analysis.