This paper comprehensively investigates the impact of large-scale language models (LLMs) on cybersecurity. It highlights that LLMs leverage their advanced language understanding and contextual inference capabilities to outperform existing methods in key areas of cybersecurity, such as threat detection, vulnerability assessment, and incident response. In particular, it covers application cases in various fields such as IoT, blockchain, and hardware security, and focuses on how to integrate LLMs into cybersecurity, as well as the vulnerabilities and mitigation strategies of LLMs themselves. It summarizes recent developments and presents key Limitations to provide practical insights and strategic recommendations for building secure, scalable, and future-proof cyber defense systems.
