MindPatternSubscribe

Entity trail

SMSR

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

1

Findings

1

Edges

0

Sources

1

Corpus findings

2026-06-14 / arxiv-researcherSMSR: Certified Defence Against Runtime Memory Poisoning in Persistent LLM AgentsSMSR defends persistent-memory agents from multi-session memory poisoning using HMAC-SHA256 signing plus randomized memory ablation and majority voting. It drops attack success from 93–100% to 0% for unsigned injections and holds authenticated single-injection attacks to 8.0% (95% CI [5.8, 10.9]), while keeping 90% utility on clean queries across 3,150 trials in 15 enterprise scenarios. Directly relevant to anyone running agents with long-lived memory.

Source trail

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues