paperSeptember 2026UnreviewedOpen access

LLM-Based Agents for Cybersecurity: A Systematic Review of Architectures, Applications, and Open Challenges

George Fatouros, Konstantinos Mavrogiorgos, Georgios Makridis, J. Soldatos, D. Kyriazis

Journal of Cybersecurity and Privacy

Abstract

The rapid evolution of Large Language Models (LLMs) has opened new frontiers in cybersecurity automation, enabling intelligent agents capable of multi-step reasoning, tool invocation, and autonomous decision-making across complex security tasks. While individual applications have emerged across threat intelligence, vulnerability assessment, penetration testing, and security operations center (SOC) automation, a systematic understanding of the LLM-based agent paradigm in cybersecurity—encompassin

Cite

@article{fatouros2026llmbased,
  title = {{LLM-Based Agents for Cybersecurity: A Systematic Review of Architectures, Applications, and Open Challenges}},
  author = {George Fatouros and Konstantinos Mavrogiorgos and Georgios Makridis and J. Soldatos and D. Kyriazis},
  year = {2026},
  month = sep,
  journal = {Journal of Cybersecurity and Privacy},
  doi = {10.3390/jcp6050159},
  url = {https://www.semanticscholar.org/paper/f0888a645148559fc788a0efb5a1c3f4cfd99d92}
}