Automating Cyber Defense: Enhancing Threat Intelligence with AI-Driven Annotation

The cybersecurity landscape is increasingly challenged by the growing digital footprint and the sophistication of cyber threats, necessitating experts to stay updated and extract actionable insights from a deluge of Cyber Threat Intelligence data. Given the impracticality of manually processing approximately 60,000 pieces of Cyber Threat Intelligence released monthly, this paper introduces the Cybersecurity Entity Extraction Tool, an efficient technique leveraging pre-trained transformer-based large language models and an artificial neural network to extract named entities from unstructured sources within the cybersecurity domain. This tool not only facilitates the identification and contextual understanding of cyber threats but also achieves a commendable F1-score of 92% across 14 distinct labels, significantly mitigating the limitations of previous methods and enhancing the capability of analysts to process large volumes of Cyber Threat Intelligence data efficiently.