Large Language Models for Real-Time Threat Detection in Cybersecurity Intelligent Systems

Security Operations Centers are drowning in alerts. Analysts simply cannot keep up, and the old signature-based tools miss too many new attacks. This paper examines whether large language models (LLMs) can actually help. It reviews recent work from AI, network security and SOC operations to see how LLMs are being used for real-time threat detection. The four main areas covered are intrusion detection, log anomaly detection, threat intelligence summarisation and automated SIEM query generation. The evidence shows that LLMs can improve detection and cut response times, especially when they are paired with retrieval-augmented generation or used in ensemble setups. But the gains come from different studies using different datasets, so direct comparison is tricky. Hallucination, adversarial attacks, privacy risks and compute costs are still big problems. The paper closes with a four-part framework for deploying LLMs responsibly and suggests where future research should go.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-12
DOI
https://doi.org/10.5281/zenodo.24151828
Primary Topic
Network Security and Intrusion Detection
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Large Language Models for Real-Time Threat Detection in Cybersecurity Intelligent Systems

Shourya Sharan
Zenodo (CERN European Organization for Nuclear Research)
Network Security and Intrusion Detection
preprint

Large Language Models for Real-Time Threat Detection in Cybersecurity Intelligent Systems

Shourya Sharan
preprint en

Abstract

Security Operations Centers are drowning in alerts. Analysts simply cannot keep up, and the old signature-based tools miss too many new attacks. This paper examines whether large language models (LLMs) can actually help. It reviews recent work from AI, network security and SOC operations to see how LLMs are being used for real-time threat detection. The four main areas covered are intrusion detection, log anomaly detection, threat intelligence summarisation and automated SIEM query generation. The evidence shows that LLMs can improve detection and cut response times, especially when they are paired with retrieval-augmented generation or used in ensemble setups. But the gains come from different studies using different datasets, so direct comparison is tricky. Hallucination, adversarial attacks, privacy risks and compute costs are still big problems. The paper closes with a four-part framework for deploying LLMs responsibly and suggests where future research should go.

Zenodo (CERN European Organization for Nuclear Research)
Network Security and Intrusion Detection
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon