Artificial Intelligence and Large Language Models in Advancing Cyber Threat Intelligence: A Systematic Literature Review

Abstract

The evolution of Cyber Threat Intelligence (CTI) is pivotal in addressing the increasingly sophisticated landscape of cyber threats. Traditional CTI methods, while foundational, are rapidly obsoleting by the complexity and dynamism of modern cyber threats. This demands a shift towards more advanced, adaptive strategies integrating Artificial Intelligence (AI) and Large Language Models (LLMs). This literature survey examines recent developments in CTI, focusing on traditional methods, the incorporation of AI for enhanced forecasting and detection, and the pioneering application of LLMs for automating intelligence report generation, enhancing threat detection, recognition, and mitigation accuracy, and facilitating the construction of comprehensive threat prevention framework. Key findings indicate that while traditional CTI methods provide a critical baseline for threat intelligence, their static nature and reliance on outdated indicators of compromise limit their effectiveness against novel, sophisticated cyber threats. Integrating AI techniques, particularly machine learning and deep learning, marks a significant advancement in CTI, offering improved predictive capabilities, dynamic threat detection, and more nuanced analysis of cyber threats. Further, the advent of LLMs in CTI represents a transformative shift, enabling the automated generation of intelligence reports, processing of unstructured CTI data, threat detection, recognition, and actionable step generation to mitigate potential cyber threats. However, challenges persist, including data privacy concerns, the dynamic nature of cyber threats, and the integration of LLMs into existing cybersecurity frameworks. The reviews provide limitations for each study and future research directions with particular emphasis on developing adaptive, intelligent CTI systems capable of proactively addressing the evolving cyber threat landscape.