Mesh Store-and-Forward Submission
Carry encrypted application payloads from an offline or intermittently-connected sender to an online relay over a delay-tolerant mesh network, with end-to-end confidentiality (intermediate peers see only ciphertext) and source-fingerprinting resistance at the radio and network layers.
Carry encrypted application payloads from an offline or intermittently-connected sender to an online relay over a delay-tolerant mesh network, with end-to-end confidentiality (intermediate peers see only ciphertext) and source-fingerprinting resistance at the radio and network layers.
The pattern is the recipient-side dual of pattern-network-anonymity.md. The umbrella pattern covers IP-layer transport (Tor, Nym, TEE-assisted); mesh store-and-forward covers the case where the sender has no persistent IP transport at all.
- Mesh transport: Bluetooth LE for short-range device-to-device, LoRa for kilometer-range, Wi-Fi Direct or local AP for higher-bandwidth corridors. Peer-to-peer or multi-hop store-and-forward.
- End-to-end encryption: IND-CCA2 authenticated encryption with ephemeral sender keying. X25519 plus ChaCha20-Poly1305 or AES-256-GCM are conventional.
- Relay decryption keypair: rotated by the relay at least every 24 hours; retired private keys securely erased.
- Source-fingerprinting mitigations: at least two orthogonal mechanisms. Physical-layer options: randomized Bluetooth LE MACs, rotated LoRa node identifiers. Network-layer: onion-routed propagation (Sphinx-style), Poisson-distributed cover traffic.
- Out-of-band relay-key distribution: relay identity public keys pre-loaded onto sender devices through trusted channels (printed cards, signed bundle, in-person handoff).
- End-to-end confidentiality: intermediate peers cannot recover the cleartext payload.
- Source unlinkability across encounters: physical-layer rotation plus network-layer cover traffic together prevent an adversary observing one mesh encounter from linking it to the same sender's prior or subsequent activity. Single-mitigation deployments do not satisfy this.
- Tolerance to high latency: delivery is eventual, bounded only by mesh path connectivity.
- Threat model: adversary observes radio spectrum and IP traffic, may control some peers, and may compromise individual companion devices. Out of scope: SIGINT-level RF fingerprinting that distinguishes individual transmitters by hardware characteristics.
- Latency. Delivery adds seconds to days, depending on path connectivity and cover-traffic schedule. Applications must tolerate this.
- Companion-device hygiene. A shared or compromised companion sees plaintext for the senders it serves. Treat companions as compromisable; rotate when practical.
- Relay-set diversity required. Senders SHOULD fan out across multiple relays with size and jurisdictional-diversity floors as deployment requirements.
- No real-time error feedback. A failed delivery may be invisible for hours. Acknowledgement tokens propagated back through the same transport cover the happy path only.
- Cover-traffic budget. Poisson cover increases sender energy cost and mesh bandwidth use; deployment tunes the rate against the threat model.
Reporters and human-rights observers in a region with periodic internet shutdowns submit encrypted reports to a journalism organization's relay using Briar over Bluetooth LE. Phones in physical proximity sync ciphertext peer-to-peer; messages traverse multiple hops until reaching a phone with online connectivity, which forwards the ciphertext to the relay over Tor. Bluetooth LE MAC randomization combined with Poisson cover-traffic intervals prevents an adversary scanning the local Bluetooth spectrum from correlating which reporter submitted what at which time.