This thesis examines the application of Generative AI, specifically Large Language Models (LLMs), to automatically generate Description Logics (DL) reasoning rules for cybersecurity ontologies.
By leveraging established cybersecurity machine-readable schemas, such as STIX (widely used for sharing threat intelligence), the LLM will analyze and understand ontology concepts, relationships, and domain constraints to propose formal reasoning rules in the style of Description Logics.
The study aims to explore how these rules can enable automated inference (deduction), allowing a system to answer questions about threats, vulnerabilities, attack patterns, and risk scenarios. In particular, we aim to identify queries and derive knowledge that analysts might be unaware of, overlook, or not consider, based on the breadth and potential of the underlying schemas and the information they contain.
Ultimately, this approach seeks to bridge the gap between unstructured cybersecurity knowledge, formal ontologies, and automated reasoning, demonstrating how LLMs can accelerate the creation of intelligent, queryable cybersecurity knowledge bases.