CAESAR decomposes LLM-based intrusion workflows into five roles with bounded coordination protocols, yielding higher success rates and lower variance than single-agent baselines on 25 CTF tasks.
G-safeguard: A topology-guided security lens and treatment on llm-based multi-agent systems
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
When LLMs Team Up: A Coordinated Attack Framework for Automated Cyber Intrusions
CAESAR decomposes LLM-based intrusion workflows into five roles with bounded coordination protocols, yielding higher success rates and lower variance than single-agent baselines on 25 CTF tasks.