MAPLE-Guard: Memory-Aware Link Enforcement Against Memory-Link Poisoning in Multi-Agent Systems
3.80T1 sourcearXiv cs.MA
arXiv cs.MAoriginal source ↗
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2608.00426.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryA defense framework, MAPLE-Guard, that places gates at the memory write, retrieval, promotion, and cross-agent reuse stages to block memory poisoning attacks in LLM-based multi-agent systems, reducing attack success rate from ~38% to under 1% on LongMemEval and AppWorld.
Why it mattersPinpoints a specific gap in current agent defenses by targeting the memory layer itself, with an open-source implementation and measured results on standard benchmarks.
Cited by
No citations on record.