paperMay 2026Unreviewed

CASPIAN: Online Detection and Attribution of Cascade Attacks in LLM Multi-Agent Systems via Cross-Channel Causal Monitoring

Kavana Venkatesh, Jafar Isbarov, Saad Amin, Murat Kantarcioglu, Jiaming Cui

Abstract

Cascade attacks in LLM multi-agent systems (MAS) arise when adversarial influence propagates across agents and leads to escalated system-level failures through complex agent interactions. Detecting such cascades is challenging, as their signals are distributed, tightly coupled across interaction channels, and often appear plausibly benign locally but may unfold quickly either within a single turn or gradually across multiple turns. Existing defenses, being largely local and text-centric, fail to

Cite

@misc{venkatesh2026caspian,
  title = {{CASPIAN: Online Detection and Attribution of Cascade Attacks in LLM Multi-Agent Systems via Cross-Channel Causal Monitoring}},
  author = {Kavana Venkatesh and Jafar Isbarov and Saad Amin and Murat Kantarcioglu and Jiaming Cui},
  year = {2026},
  month = may,
  eprint = {2605.19240},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.19240}
}