Ontology Learning for Hybrid Threats
This technical report discusses the potential use of advanced Artificial Intelligence technologies for mining Open Source Information in the context of Hybrid Threat analysis. Increasing international tensions depict a scenario where Hybrid Threats acquire even more importance than in past years and monitoring such activities poses important technological challenges. In this scenario, Large Language Models show promising results but also the corresponding fundamental limitations. Hallucinations, biases, and lack of knowledge are emerging as issues difficult to tackle, especially in the field of Hybrid Threat analysis. We propose the use of formal knowledge bases to correct Large Language Model issues. Despite their interesting applications knowledge bases require humanintensive labor. After exploring the state-of-the-art we propose a methodology to reduce the human effort and learn automatically part of the Knowledge base. This last step involves again Large Language Models but, with the precaution of exploiting only a simple restricted set of linguistic abilities, we avoid the common pitfalls observed before. We expect to enhance the quality of Open Source Intelligence by having more reliable analysis of textual data.
BOSSO Francesco;
RUBERTO Stefano;
2025-01-31
Publications Office of the European Union
JRC140863
978-92-68-23985-8 (online),
1831-9424 (online),
EUR 40199,
OP KJ-01-25-042-EN-N (online),
https://publications.jrc.ec.europa.eu/repository/handle/JRC140863,
10.2760/8060534 (online),
Additional supporting files
| File name | Description | File type | |
|---|
