Tool Use Enables Undetectable Steganography in Multi-Agent LLM Systems

arXiv 2606.28425•96212aa983e3b5a958c808c8aa7b62ba2c49e05e12b3b6af2ee138f5acd11c99
AgentThreadHMAC pseudonymizationLLM agentsLiDAR poisoningNIDS evasionOCR/audio injectionROS 2ScopeGateTRiSMadversarial trafficagent coordinationauthorizationcapability gatingconfused-deputycovert channelsdifferential privacyformal security analysishealthcare agentsmemorizationprivacy-preserving fine-tuningprompt injectionprotocol verificationrobotics securitysensor spoofingsteganography

Paper metadata

arXiv ID
2606.28425
Version
Not specified by this published record
Category
Computer Science — Cryptography and Security (cs.CR)

The PDF link points to arxiv.org. Baitaphish does not expose a private stored PDF.

Evidence and limitations

Source ID
arxiv_cs_cr
Record identifier
96212aa983e3b5a958c808c8aa7b62ba2c49e05e12b3b6af2ee138f5acd11c99
Enrichment time
2026-06-30T07:23:29Z
AI-assisted enrichment
Yes

This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.