Dynamic Cyber Ranges with LLM defender agents reduce attacker success to 0-55% and preserve evaluation headroom as models advance by using comparable capabilities on both sides.
Cai fluency: A framework for cybersecurity ai fluency.arXiv e-prints, pages arXiv–2508
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Dynamic Cyber Ranges
Dynamic Cyber Ranges with LLM defender agents reduce attacker success to 0-55% and preserve evaluation headroom as models advance by using comparable capabilities on both sides.