The DOI matches its Crossref record; 13 of the 13 references it cites were found in a registry or an index.
iThelma: An Autonomous LLM Agent for Cyber Threat Hunting via Playbook-Driven Intelligence
Abstract
We present an autonomous agent for cyber threat hunting that integrates structured knowledge from human-authored playbooks with the generative and reasoning capabilities of large language models (LLMs). Building on a prior LLM-based threat-hunting framework, this work introduces additional components for playbook ingestion, hunt script validation, co-occurrence modeling, prompt-based generation, and tuning-based adaptation. The agent generates, validates and refines hunt scripts in a sandboxed environment, identifies behaviorally consistent scripts using consensus voting, and constructs a threat co-occurrence matrix to inform hunt scheduling. Unlike the earlier system that relied solely on natural language prompting, our design enables the agent to learn from execution feedback and adapt its models over time. This work contributes a scalable, self-adaptive and resilient foundation for LLM-driven cyber threat hunting systems.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“These capabilities are coordinated by control loops, the most influential of which are ReAct (Yao S. et al., ICLR 2023 [89]), which interleaves natural-language reasoning with tool invocations [25]; plan-and-solve, which produces a complete plan before execution; and reflection, which critiques prior outputs before committing [28,32].”
From Agentic AI and Large Language Models for Autonomous IoT Cybersecurity: A Systematic Survey, Taxonomy, and Research Roadmap (2026) · cites this paper as [25]
Citations
Cites
Code Llama: Open Foundation Models for Code
Rozière, B., Gehring, J., Gloeckle, F., Sootla, S., Gat, I., Tan, X. E., Adi, Y., Liu, J., Sauvestre, R., Remez, T., et al. (2023) · arXiv · doi:10.48550/arxiv.2308.12950 · Cited by 531 · cited 1×
“Models like Meta’s Code Llama [13], OpenAI’s GPT-4 [14], and Google’s Gemini [15] generate human-like text and executable code [16], enabling hunt script generation from natural language descriptions.”
When LLMs meet cybersecurity: a systematic literature review
Zhang, J., Bu, H., Wen, H., Liu, Y., Fei, H., Xi, R., Li, L., Yang, Y., Zhu, H., & Meng, D. (2025) · Cybersecurity · doi:10.1186/s42400-025-00361-w · Cited by 383
A survey of large language models for cyber threat detection
Chen, Y., Cui, M., Wang, D., Cao, Y., Yang, P., Jiang, B., Lü, Z., & Liu, B. (2024) · Computers & Security · doi:10.1016/j.cose.2024.104016 · Cited by 172
Enabling Efficient Cyber Threat Hunting With Cyber Threat Intelligence
Gao, P., Shao, F., Liu, X., Xiao, X., Qin, Z., Xu, F., Mittal, P., Kulkarni, S. R., & Song, D. (2021) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde51399.2021.00024 · Cited by 164
A Survey on Threat Hunting in Enterprise Networks
Nour, B., Pourzandi, M., & Debbabi, M. (2023) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2023.3299519 · Cited by 99
Gemini-the most powerful LLM: Myth or Truth
Islam, R., & Ahmed, I. (2024) · Information Communication Technologies Conference (ICTC) · doi:10.1109/ictc61510.2024.10602253 · Cited by 98
LLMCloudHunter: Harnessing LLMs for Automated Extraction of Detection Rules from Cloud-Based CTI
Schwartz, Y., Ben-Shimol, L., Mimran, D., Elovici, Y., & Shabtai, A. (2025) · International World Wide Web Conference (WWW) · doi:10.1145/3696410.3714798 · Cited by 34
What if GPT4 Became Autonomous: The Auto-GPT Project and Use Cases
Fırat, M., & Kuleli, S. (2023) · Journal of Emerging Computer Technologies · doi:10.57020/ject.1297961 · Cited by 29
Malicious Code Detection Using LLM
Hossain, A. A., Mithun Kumar, P. K., Zhang, J., et al. (2024) · IEEE National Aerospace and Electronics Conference (NAECON) · doi:10.1109/naecon61878.2024.10670668 · Cited by 12
LLMs for Code: The Potential, Prospects, and Problems
Sharma, T. (2024) · IEEE International Conference on Software Architecture Companion (ICSA-C) · doi:10.1109/icsa-c63560.2024.00067 · Cited by 11
Research works printed but not linked: 7
- Karuna, 2021, Automating cyber threat hunting using nlp, automated query generation, and genetic perturbation, arXiv preprint
- Forge, 2022, The threat hunter playbook
- 2023, Threat hunting and detection
- 2024, Gpt-4 technical report
- Bert-Jan, 2024, Kql sentinel & defender queries
- Strom, 2018, Mitre att&ck: Design and philosophy
- Rodriguez, 2022, Security datasets
Cited by
Sorted by citations
Nageshwaran, V., & Ezekiel, S. (2026) · Electronics · doi:10.3390/electronics15122740 · Cited by 4
“SOAR-LLM [28], Agentic AI for threat hunting [24,25], and the SOC hyper-automation architecture of Alzughaibi et al. [28] follow this pattern.”
InfluentialLiu, J., Lin, R., Lo, S., et al. (2026) · IEEE Conference on Communications and Network Security · doi:10.1109/cns69863.2026.11720931
“Thelma [23] and its successor iThelma [24] introduced LLM agents that ingest threat-intelligence playbooks, generate hunt scripts in a sandbox, and refine them through execution feedback.”
Self-citation
Metadata
- Authors
- Chen, N., Lin, R., Xie, D., Lin, H., & Chen, S.
- Affiliations
- National Taiwan University of Science and Technology (TW)
- Venue
- 2025 IEEE Conference on Communications and Network Security (CNS)
- Publisher
- IEEE
- Published
- 2025-09-08
- Pages
- 1-6
- ISSN
- 2474-025X (ISSN-L)
- OpenAlex
- W7091667325
- Semantic Scholar
- 551b54762eab2499e4e74024556e00ea9070235f
- Cited by
- 2
- References
- 12
- Citations by year
- 2026: 2
- FWCI
- 2.49
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- IEEE Conference on Communications and Network Security (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › Network Security and Intrusion Detection
- Physical Sciences › Computer Science › Information Systems › Digital and Cyber Forensics
- Physical Sciences › Computer Science › Information Systems › Information and Cyber Security
- Keywords
- cyber threat hunting, large language models, autonomous agents, sandboxed execution, consensus voting, self-adaptive systems, prompt-based generation
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
