Salami Attack: Stealthy Collusive Memory Poisoning against OpenClaw

Abstract

Long-term memory enables LLM agents to retain useful information across sessions, but also creates an attack surface through which adversaries may poison an agent's persistent memory to steer its behavior. Existing memory poisoning attacks mainly rely on individually malicious records, overlooking a compositional threat: multiple benign-looking memories may jointly induce unsafe behavior. In this paper, we introduce MemCollusion, an automated red-teaming framework for constructing collusive memo

Categories

Framework mappings

Suggested from the entry's categories.

@misc{lin2026salami,
  title = {{Salami Attack: Stealthy Collusive Memory Poisoning against OpenClaw}},
  author = {Zheng Lin and Yuzhen Huang and Zhenxing Niu and Xianmin Ye and Haichang Gao},
  year = {2026},
  month = aug,
  eprint = {2608.01637},
  archivePrefix = {arXiv},
  url = {https://www.semanticscholar.org/paper/78a6fa7c6a59bb5fe8b4337b6f1b3d9b7b69328b}
}