A LARGE LANGUAGE MODEL-BASED SECURITY OPERATIONS DECISION-SUPPORT AGENT FOR ENTERPRISE CYBERSECURITY: ARCHITECTURE, RISK-AWARE COORDINATION, AND SENSITIVITY TO PARTIAL LOG TOKEN MASKING
This study evaluates an LLM-based security-operations decision-support architecture using a context-matched rule baseline, knowledge augmentation, and a policy-conditioned recommendation layer. On AIT-LDS, the LLM substantially improves detection over the tested deterministic baseline, while knowledge and recommendation conditioning reduce false positives. Robustness is examined through deterministic token masking and dependence-aware analysis.
Сведения
- Опубликовано
- 2026 · Т. 4, № 10, С. 22-35
- Язык
- Английский
- ISSN
- 3030-3702
- Издатель
- LLC Scienceproblems Team