Security
Veracium is built so that exposure of any single component does not expose the people on the ground.
Architecture in one sentence
The system is a blind intermediary: the Journalist* knows who they are, the editor knows what they received, and Veracium operates the layer in between without ever holding both halves in the clear.
How it holds up
- On-device sealing — Hashing happens before transmission. We cannot tamper with what we never see in raw form.
- Hardware-backed signing — Keys live in the device's secure enclave and cannot be extracted, even by Veracium.
- Shamir Secret Sharing on identity — The link between Journalist pseudonym and real identity is split across multiple custodians. No single party — including Veracium — can reconstruct it alone.
- zk-SNARKs for tier proofs — Editors can verify that a Journalist holds the required tier without seeing who they are.
- EU-hosted, EU-jurisdiction — All production data lives on PostgreSQL in Frankfurt. No US data transfer for operational records.
What we do not protect against
We are explicit about limits. Veracium does not protect against:
- Coerced capture by a hostile party present at the scene.
- An Journalist deliberately staging a scene that is then truthfully sealed.
- A editor republishing the capture without the accompanying certificate.
For each of these, the certificate and chain-of-custody log are designed to make the problem detectable after the fact even if not preventable in the moment.
Reporting a vulnerability
jorg@veracium.io · PGP key fingerprint on /about. We respond within 72 hours and credit responsible disclosure on the project page.