Scholarlink

Cyber threats are becoming increasingly complex, causing traditional security systems to struggle in keeping up and highlighting the need for advanced solutions. Large Language Models (LLMs), such as OpenAI’s ChatGPT and Meta AI’s LLaMA, have shown great potential to transform cybersecurity workflows with their abilities in natural language understanding, pattern recognition, and automated reasoning. These models are particularly promising for tasks like network monitoring, threat detection, and security alert triage. However, challenges related to the reliability of outputs, adversarial risks, and ethical concerns must be addressed. This paper presents a comprehensive survey of LLM-based approaches for security testing and evaluates three open-access LLMs, including Mistral-7B, Qwen3-8B, and Llama3.1-8B, demonstrating their ability to enhance security alert analysis. Our findings suggest that LLMs can improve alert clarity and usability, making them more accessible to non-experts while providing valuable insights for developers.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

LLMs in Security Testing and Monitoring: An Initial Study

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Luong Nguyen,
  • Manh-Dung Nguyen,
  • Anh Hao Bui,
  • Ana Rosa Cavalli,
  • Edgardo Montes de Oca

摘要

Cyber threats are becoming increasingly complex, causing traditional security systems to struggle in keeping up and highlighting the need for advanced solutions. Large Language Models (LLMs), such as OpenAI’s ChatGPT and Meta AI’s LLaMA, have shown great potential to transform cybersecurity workflows with their abilities in natural language understanding, pattern recognition, and automated reasoning. These models are particularly promising for tasks like network monitoring, threat detection, and security alert triage. However, challenges related to the reliability of outputs, adversarial risks, and ethical concerns must be addressed. This paper presents a comprehensive survey of LLM-based approaches for security testing and evaluates three open-access LLMs, including Mistral-7B, Qwen3-8B, and Llama3.1-8B, demonstrating their ability to enhance security alert analysis. Our findings suggest that LLMs can improve alert clarity and usability, making them more accessible to non-experts while providing valuable insights for developers.