A search-based simulation method uses LLMs as optimizers to discover escalating privacy attacks like impersonation and corresponding defenses like identity-verification state machines in LLM agent dialogues.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Searching for Privacy Risks in LLM Agents via Simulation
A search-based simulation method uses LLMs as optimizers to discover escalating privacy attacks like impersonation and corresponding defenses like identity-verification state machines in LLM agent dialogues.