Toward Human-Aware Active Directory Defense with Fine-Tuned LLMs
Proceedings of the AAAI Symposium Series · 23.06.2026 · 9 (1) · 332-335
DOI: 10.1609/aaaiss.v9i1.42949 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Symposium Series, online erschienen am 23.06.2026. Verfasst von Hung X. Nguyen und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Autonomous cyber defense agents increasingly need to query complex enterprise attack graphs under time pressure, yet today’s Active Directory (AD) tools still assume expert-authored Cypher queries. This paper studies natural-language-to-Cypher generation as a core tool-use action for human-aware AD defense agents operating over BloodHound-style graphs. We curate 346 executable English–Cypher pairs from practitioner queries, augment them to 2,768 samples via constrained paraphrasing, and fine-tune an open-weight Mixtral‑8×7B model using QLoRA. Across ten representative BloodHound-equivalency tasks on five synthetic AD graphs, fine-tuning raises parse success from 0.80 to 0.94 and correct-answer rate from 0.34 to 0.42. These gains should be interpreted as evidence that domain tuning can improve syntactic robustness in a controlled setting, rather than as evidence of production-ready semantic reliability or real-enterprise scalability. Error analysis reveals safety-critical semantic gaps in domain admin identification, unsupported-OS detection, and temporal constraints. We argue that, for cyber defense agents, schema prompting and fine-tuning alone are insufficient: safe autonomy over AD graphs requires execution-grounded correction, explicit domain guardrails, and interaction designs that calibrate human trust in LLM-generated queries.
Zitieren
APA
Nguyen, H. X., & Vu, T. (2026). Toward Human-Aware Active Directory Defense with Fine-Tuned LLMs. Proceedings of the AAAI Symposium Series, 9(1), 332-335. https://doi.org/10.1609/aaaiss.v9i1.42949
Vancouver
Nguyen HX, Vu T. Toward Human-Aware Active Directory Defense with Fine-Tuned LLMs. Proceedings of the AAAI Symposium Series. 2026;9(1):332-335. doi:10.1609/aaaiss.v9i1.42949
BibTeX
@article{Nguyen2026,
author = {Nguyen, Hung X. and Vu, Tu},
title = {Toward Human-Aware Active Directory Defense with Fine-Tuned LLMs},
journal = {Proceedings of the AAAI Symposium Series},
year = {2026},
volume = {9},
number = {1},
pages = {332--335},
doi = {10.1609/aaaiss.v9i1.42949}
}
EUVIMED