.jpg)
In the current cybersecurity landscape, threat intelligence (CTI) has become a fundamental pillar for anticipating attacks and protecting critical infrastructures. However, most CTI reports arrive in unstructured formats, full of noise and heterogeneity, which hinders their automated processing. To address this challenge, cybersecurity knowledge graphs (CSKG) have emerged, organizing entities, actions, and adversarial relationships into a navigable structure. But building these graphs from free text remains a complex task, especially when using monolithic large language models (LLMs) that attempt to perform extraction, typing, and verification all at once, generating high costs, low controllability, and unstable results.
Faced with this problem, an innovative proposal is TACTIC-KG: a framework based on specialized agents that breaks down CSKG construction into modular tasks. Instead of a single giant LLM, several lightweight agents (3B to 8B parameters) are used to extract entities, assign types, verify consistency, and curate data. This architecture not only improves the stability, recall, and coherence of the graph but also drastically reduces deployment costs. Experiments show that specialized agents outperform monolithic in-context learning models in F1-score, typing accuracy, and structural similarity.
This trend towards the specialization of artificial intelligence agents for businesses reflects a paradigm shift: the goal is no longer a single all-powerful model, but an ecosystem of small assistants that collaborate. At Q2BSTUDIO, we understand that modern cybersecurity requires flexible, scalable solutions tailored to each business. Therefore, we offer custom applications that integrate AI agents to automate threat analysis processes, as well as AI for businesses that enhances business intelligence. Our team develops custom software that can orchestrate multiple lightweight agents, similar to those in TACTIC-KG, to build personalized knowledge graphs.
Furthermore, the implementation of these systems benefits from a robust cloud infrastructure. From AWS and Azure cloud services, at Q2BSTUDIO we help deploy modular architectures that optimize costs and latency, allowing even small AI agents to work in parallel. Cybersecurity cannot afford bottlenecks, and with specialized agents, a perfect balance between performance and economy is achieved. We also integrate business intelligence services with tools like Power BI to visualize relationships extracted from graphs, transforming threat data into actionable dashboards.
In summary, TACTIC-KG demonstrates that the key is not the size of the model, but the intelligence of the design. Divide and conquer also works in cyber knowledge extraction. For companies looking to implement advanced cybersecurity solutions, having a technology partner that offers both custom software and expertise in AI agents and cloud is essential. At Q2BSTUDIO, we transform these concepts into operational realities, helping organizations anticipate threats with precise and efficient intelligence graphs.
