What we know
Formal analysis of honeytoken limits when trusted agents and attackers share memory or policies.
Evidence limits
Results are theoretical and depend on assumptions about knowledge, repetition and response laws.
Reviewed: Full text.
Full-text review · 2026-09-16
Design and scope
Formal analysis using total-variation bounds and repeated learning under shared memory.
Main finding
Shows that policies observable to trusted agents can leak how to distinguish honeytokens; proposes a private reference monitor.
Evidence limits
Results are theoretical and depend on assumptions about knowledge, repetition and response laws.
Sources and provenance
- When Agents Talk: Honeytokens under Shared Memory
Full text · 2026-09-16
Reviewed: 2026-09-16. This record may change when new evidence is found.