Logo image
Back
A Survey of Large Language Models (LLMs) for Cybersecurity: Opportunities and Directions
Conference proceeding   Peer reviewed

A Survey of Large Language Models (LLMs) for Cybersecurity: Opportunities and Directions

Md Abdur Rahman, Guillermo Francia, Hossain Shahriar, Alfredo Cuzzocrea, Atef Mohamed, Muhammad Umair Khan and Sheikh Iqbal Ahamed
IEEE International Conference on Big Data, pp.4333-4342
International Conference on Big Data (BigData) (Macau, China, 12/08/2025–12/11/2025)
12/08/2025
DOI: https://doi.org/10.1109/BigData66926.2025.11402639

Metrics

1 Record Views

Abstract

The recent development of Large Language Models (LLMs) has shown remarkable success in data-driven applications. Trained on massive textual datasets, LLMs have demonstrated the ability to serve various tasks in cybersecurity. More importantly, the power of Transformers brings us closer to human-like performance in data-centric applications. These capabilities can be leveraged to detect cyber threats, mitigate risks, and prevent attacks. Researchers demonstrate gaps in this area to apply LLMs as they may have misunderstanding to generate text dataset for it. In this work, we provide an overview of recent applications of LLMs to handle data scarcity, use reduced number of observation to train Machine Learning (ML) models or fine-tune LLMs, or integrate LLMs to ML models to improve existing ML models. Also, we explore the papers that leverages various lLMs to use non-text dataset for cyberattacks detection. Finally, we discuss the recent works about how LLMs leverage to transform traditional text or non-text dataset to vectorize forms for training various ML models or fine-tune various Bidirectional Encoder Representations from Transformers (BERT) to serve different purposes. We also discuss the challenges and future directions of LLM research in cybersecurity. Furthermore, we survey recent studies in each area, highlighting their strengths and weaknesses. Finally, this outline potential future research directions for maximizing the benefits of LLMs in cybersecurity.

Details

Logo image