Large Language Models (LLMs) hold significant promises to revolutionize cybersecurity. Unlike traditional machine learning (ML) models, LLMs can be fine-tuned for specific tasks such as identifying malicious code, analyzing software vulnerabilities, and detecting phishing attacks. However, challenges remain, including the high computational cost of development, potential biases in training data, and the risk of adversarial attacks against these models. This review examines the practicality of using LLMs in cybersecurity with current technologies over the next five years. It evaluates their real-world applicability and draws on recent literature to highlight potential security threats, benefits, and implementations of LLMs in this domain. These studies provide examples where LLM-based tools have been both effective and flawed, helping establish precedents for LLM use. Additionally, this review discusses the history and rapid development of LLMs, comparing current advances to past technological growth, and explores future research directions for integrating LLMs into cybersecurity.
Illindala et al. (Wed,) studied this question.