This is an architectural prototype, not a fielded product, and the distinction matters. A vendor who cannot tell you where their architecture hurts has not looked.
| Component | State | Detail |
|---|---|---|
| ChaCha20-Poly1305, HKDF, Ed25519 | Real · KAT | Known-answer tested against RFC 8439 §2.3.2, §2.5.2 and §2.8.2, RFC 5869 and RFC 8032 §7.1, and cross-checked against an independent implementation. Correctness-oriented, not constant-time, not side-channel hardened. |
| Shamir over GF(2⁸) + proactive refresh | Real · tested | Any t of n reconstruct; t−1 leave every secret value equally possible; cross-epoch shares yield garbage. |
| Reed–Solomon MDS coding | Real · 300 subsets | Every k-subset decodes, every (k−1)-subset refuses. Used for availability only; confidentiality below threshold comes from the shared key. |
| Dispersal (seal, code, share the key) | Real · tested | Verified in the browser reference engine: any k shards recover and authenticate, a corrupted shard is detected, and the scheduler never places k shards on one bearer. |
| Hybrid KEM combiner | Real | The security-critical part: both ciphertexts bound into the KDF, so the session key is secure if either component is. Unchanged when backends are swapped for certified modules. |
| Release policy tree, UAS ledger | Real · tested | Linear secret sharing over the access structure with fixed-shape padding; hash-chained signed provenance with tamper detection. |
| Browser reference engine (this site) | Real · 25 tests | Everything in the lab runs on it, client-side, with no network calls. Built for explanation, not for protecting traffic. |
| ML-KEM-1024 | Sized stub | No post-quantum security whatsoever yet. Buffer lengths match FIPS 203 so MTU budgeting and shard sizing are honest. Swapped for a certified module at integration; nothing above it changes. |
| Module validation, formal verification | Not started | Needs a validated module and an evaluation laboratory engagement through the national approval route. Tamarin or ProVerif models of the quorum unlock and the hybrid handshake are planned. |
| Hardware integration, accreditation | Not started | The work that turns any of this into a fielded system. |
The prototype is a specification you can execute, not a shipping crypto library. It is dependency-free and runs on an air-gapped machine, which is where a defence customer will want to run it. Prototype and test vectors are available on request.
Quorum identity has a real cost for the genuinely cut-off operator, and a radio cannot tell isolation from capture. The grace-period mitigation is a policy dial, not an answer, and it needs to be argued out in a user trial rather than defaulted.
4.8× emission overhead at full cover is survivable on a vehicle and painful on a dismounted radio over 72 hours. The adaptive budget helps; it does not eliminate the problem, and it must not track tempo.
Assembling a quorum for every signing operation is fine at message cadence and wrong at packet cadence. Caching a quorum-derived session key weakens the capture property in proportion to the cache lifetime. That trade is unresolved.
The symmetric construction is right for a single national authority and wrong for genuine coalition federation.
Enumerating available bearers is itself an observable act on some of them.
Every one of these is a question we want answered in a user trial with a real signals regiment, which is what the iDEX route is for.
| Phase | Deliverable | Evidence it worked | Kill criterion |
|---|---|---|---|
| 0–6 months | Certified crypto core: ML-KEM / ML-DSA through a validated module; Tamarin model of quorum unlock and handshake. | Formal proofs discharge; known-answer tests pass against a validated module. | Quorum latency cannot be made workable at message cadence. |
| 6–12 months | Saelvor on ruggedised COTS. Lab trial against a real EW team. | Traffic-analysis correlation stays near zero under live collection; emission profile within the agreed DF budget. | Cover-traffic power cost is unsurvivable on a 72-hour dismounted battery. |
| 12–18 months | iDEX user trial with a signals regiment. Quorum Lace / UAS with a UAS unit. | Operators complete tasks without workarounds; the isolated-node problem is bounded. | Operators route around quorum identity in the field. |
| 18–30 months | Accreditation pathway; Release coalition pilot; first Chapter III order. | Signed order. Second customer without a founder in the room. | No accreditation route exists for software-first COMSEC at the target classification. |
Each kill criterion is a real one. We would rather find them in a trial than in service.
Not capital. If you can open that door, or want the architecture note and the runnable prototype to take apart yourself, we would like to hear from you.