This research introduces the Communication Inference Attack (CIA), a novel method exposing privacy risks in LLM-based multi-agent systems by reconstructing i...
Level: advanced
By Yongxuan Wu
Category: discussion