Devices that prove where they are, and a ledger that keeps the record
The system runs from firmware to chain, in two-week sprints with acceptance criteria the client signs off. The current phase brings a public test network and block explorer.
Stack and topics
| Industry | Deep tech, connected devices |
|---|---|
| Region | Europe |
Context
The client and the problem
A venture building proof of physical presence needed radios, sensors and phones to produce evidence that a person or device was at a place at a time, signed so that nobody can fake it later, with a blockchain keeping the record.
That is a hard problem from both ends. At the device end, radio ranging is noisy and every unit behaves a little differently. At the ledger end, the rewards and penalties around a proof only work if the proofs themselves cannot be relayed, replayed or produced by a farm of fake witnesses.
What we built
What did we build?
We built the system end to end, from firmware on the device to the chain and the reward engine, in our own hardware lab.
- Firmware in Rust. Firmware for ESP32-C6 with Wi-Fi FTM ranging and a seven-step accuracy pipeline, live tuning over MQTT, per-device calibration and over-the-air updates.
- Radio and sensing research. Wi-Fi RTT, UWB, BLE, LTE, mmWave radar and Wi-Fi CSI, each tested against a written acceptance plan, plus the hardware specification and procurement.
- Signed multi-party proofs. Challenge and response with witness quorums and anti-collusion rules, Ed25519 signatures on every message, an off-chain proof index and a trust grade on every proof.
- Threat model first. Relay and wormhole attacks, Sybil witness farms, replay and clock skew were each mapped to a control before code was written.
- Ledger and incentives. A chain chosen across twelve candidates, a private proof-of-stake Ethereum network, token design with staking, slashing and reputation, and a deterministic reward engine.
- Machine learning on sensor data. Pose and activity models trained on mmWave point clouds, and occupancy heatmaps.
Protocol
How does the proof work?
The verifier sends a challenge with a nonce. The device signs the nonce together with its ranging data using Ed25519 and returns the signed response. The verifier asks a witness quorum to confirm; the proof counts once the quorum agrees under the anti-collusion rules. The proof is indexed off-chain with a trust grade and the ledger keeps the record.
How we worked
Team and process
The work runs in two-week sprints. Every sprint has acceptance criteria that the client signs off, and every radio or sensing option is tested against a written acceptance plan before it goes into the design.
Engineering standards are part of the scope: Rust 2024 with no unsafe code in shared crates, rustls, OpenTelemetry tracing, cargo-audit, cargo-deny, SBOMs and idempotency keys. These are the same controls that the EU Cyber Resilience Act will expect from connected products.
Facts
Project facts
| Fact | Detail |
|---|---|
| Client | A deep-tech venture in connected devices |
| Region | Europe |
| Our role | Firmware, radio research, protocol, threat model, chain and incentives |
| Stack | Rust, ESP32-C6, Wi-Fi FTM, MQTT, Ed25519, private proof-of-stake Ethereum network |
| Process | Two-week sprints, acceptance criteria signed by the client |
Result
Where it stands
The system runs from firmware to chain: devices produce signed proofs, witnesses confirm them, and the ledger records them and settles the rewards.
The current phase brings a public test network and a block explorer.
Next step
Building something similar?
A 30 minute call with Oleg. You leave with a clear view of scope, risks and the next step.
Oleg Pun
CEO, Vareger
