LLMs and ontologies combine for transparent cyber threat intelligence

Researchers have developed a new method for cyber threat intelligence that combines Large Language Models (LLMs) with domain ontologies. This approach aims to improve the accuracy and explainability of extracting information from cybersecurity logs, particularly for unstructured or ambiguous entries. By integrating ontologies and constraints, the AI agent structures extracted data into an enriched graph database, enhancing semantic analysis capabilities. Evaluations using public datasets showed this method outperforms traditional prompt-only techniques in extraction quality. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Enhances accuracy and explainability in cyber threat intelligence, potentially improving security operations.

RANK_REASON This is a research paper detailing a novel methodology for cyber threat intelligence.

Read on arXiv cs.AI →

• LLM
• AI
• Luca Cotti
• SHACL
• Hugging Face
• arXiv

• paper
• product