Searching for Diamonds: Cross-Domain Opportunities in Cyber Threat Intelligence
Open Access
- 30 October 2025
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 189554-189588
Abstract
Natural Language Processing (NLP) and Large Language Models (LLMs) are increasingly used in cybersecurity to enable automated, auditable, and intelligent systems. However, this convergence remains conceptually and methodologically underdeveloped in Cyber Threat Intelligence (CTI), a domain centered on processing large volumes of information. This paper presents a cross-domain review that synthesizes the state-of-the-art literature and outlines strategic opportunities, risks, and open challenges related to the application of LLMs in CTI. Guided by the conceptual lens of “searching for diamonds,” the study identifies cross-cutting aspects—such as explainability, semantic reasoning, federated collaboration, model security, and machine unlearning—as essential to building normatively aligned autonomous CTI agents capable of efficiently processing the high-volume, dynamic data characteristic of CTI environments. It also highlights gaps in validation, trust, and regulatory compliance, emphasizing the need for co-evolutionary defenses and formal verifiability mechanisms. Grounded in multidisciplinary insights and aligned with international frameworks such as the Tallinn Manual and NIST/ISO standards, this work provides a strategic roadmap for future research and deployment of LLM-driven CTI architectures.Keywords
Funding Information
- Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), National Council for Scientific and Technological Development
- Programa de Pós-Graduação em Aplicações Operacionais (PPGAO), Aeronautics Institute of Technology
- São Paulo Research Foundation (#2020/09850-0, #2022/00741-0)
- C2-DC Laboratory for providing the infrastructure used in the experiments
This publication has 133 references indexed in Scilit:
- Is text preprocessing still worth the time? A comparative survey on the influence of popular preprocessing methods on Transformers and traditional classifiersInformation Systems, 2023
- Modelling direct messaging networks with multiple recipients for cyber deception2022
- SoK: The Impact of Unlabelled Data in Cyberthreat Detection2022
- Trojaning Language Models for Fun and Profit2021
- FastSpec: Scalable Generation and Detection of Spectre Gadgets Using Neural Embeddings2021
- Fall of Giants: How popular text-based MLaaS fall against a simple evasion attack2021
- Extractor: Extracting Attack Behavior from Threat Reports2021
- Jekyll: Attacking Medical Image Diagnostics using Deep Generative Models2020
- ChainSmith: Automatically Learning the Semantics of Malicious Campaigns by Mining Threat Intelligence Reports2018
- Forgotten Siblings: Unifying Attacks on Machine Learning and Digital Watermarking2018