Post-quantum authentication
Post-quantum cryptography protects two different things. Secrecy means a call recorded today stays unreadable to someone who builds a quantum computer later. Authentication means that same quantum computer can’t pretend to be someone it isn’t. MOQOM is post-quantum for both.
Secrecy
Section titled “Secrecy”Every key that protects content is agreed or sealed with post-quantum cryptography:
- Room keys. Devices agree each room’s keys with MLS
in the cipher suite
MLS_256_MLKEM1024_AES256GCM_SHA512_MLDSA87, the only suite, whose key agreement is ML-KEM-1024. Every part of it is NIST category 5. It is the category-5 suite of draft-ietf-mls-pq-ciphersuites; see moqom-mls. - Media. SFrame encrypts every frame with AES-256-GCM. A 256-bit symmetric key keeps about 128 bits of security against a quantum computer.
- Large broadcasts. With sender keys, each viewer’s device makes an ML-KEM-1024 key pair, and each epoch’s key is sealed to it with HPKE over ML-KEM-1024.
- Connections. TLS 1.3 runs on every connection,
and its key exchange is ML-KEM-1024 only, on every device connection (media, sign-in and token
renewal, MLS key delivery) and on every link between MOQOM’s servers. The Swift SDK connects
through MOQOM’s own QUIC and TLS library (rustls with AWS-LC), so the SDK chooses the key
exchange. Classical-only key exchange is refused everywhere. See
End-to-end encryption for
api.moqom.cloud, which browsers also reach.
Authentication
Section titled “Authentication”Every signature that says who someone is uses ML-DSA-87 (FIPS 204):
| What’s signed | Algorithm |
|---|---|
| Each device’s identity key, generated in the Secure Enclave and never exportable | ML-DSA-87 |
| The device credential: the proof of possession every relay demands | ML-DSA-87 |
| Every MLS KeyPackage | ML-DSA-87 |
| Every MLS commit | ML-DSA-87 |
| Every MLS message signature, handshake and application messages alike | ML-DSA-87 |
| Direct tickets issued by relays | ML-DSA-87 |
| Join tokens, signed and then sealed with AES-256-GCM | ML-DSA-87 |
| The room’s control log | ML-DSA-87 |
There is no classical signature algorithm anywhere in that list, and none is accepted. Relays refuse ES256 proofs, verifiers refuse Ed25519 tokens, and a key set that contains an elliptic curve key is refused when it is loaded.
Why authentication matters
Section titled “Why authentication matters”A quantum computer that can break elliptic curves can work out a private key from its public key. Public keys are public by design: they’re published, exchanged and compared in safety numbers. Against classical identity keys, that quantum computer could impersonate a contact or forge identity keys in new sessions. The safety number would still match, because the public key is the real one; the attacker simply holds its private half too.
Post-quantum secrecy alone doesn’t stop this. It keeps recordings unreadable, but an active attacker who can sign as you can join a new conversation as you. With ML-DSA-87 identity keys, impersonation is quantum-resistant too.
Compared with Signal
Section titled “Compared with Signal”Signal added post-quantum key agreement with PQXDH in 2023 and its post-quantum ratchet, SPQR, in 2025, so recorded Signal messages stay secret against a future quantum computer. Signal’s identity keys and signatures are Curve25519 and XEdDSA, which are classical. Signal’s own PQXDH specification states that its authentication relies on classical assumptions and isn’t secure against an active quantum attacker. So a future quantum computer couldn’t read recorded Signal messages, but it could impersonate a contact or forge identity keys.
| What’s signed | MOQOM | Signal |
|---|---|---|
| Recorded messages stay secret | Yes | Yes |
| Long-term identity key | ML-DSA-87 | Curve25519 |
| Prekeys or KeyPackages signed with | ML-DSA-87 | XEdDSA |
| Impersonation by a future quantum computer | Resisted | Possible in new sessions |
Signal column: our reading of Signal’s published specifications as of October 2026.
What is still classical
Section titled “What is still classical”Two signatures in the path are made by others with classical algorithms. Neither can sign as an existing device.
- Certificate signatures in TLS. Relay certificates, and the Let’s Encrypt certificates on MOQOM’s public endpoints, are ECDSA or RSA, because the WebPKI has no post-quantum certificates yet. The key exchange is post-quantum; the certificate’s signature is not. Phones accept only the exact relay certificate they were given, every proof is bound to that certificate, and media is end-to-end encrypted regardless. Forging the handshake would need a working quantum computer during the connection, and would get ciphertext.
- Apple’s sign-in. Sign in with Apple identity tokens and App Attest are signed by Apple with classical algorithms. Someone who could forge one could at most enroll a new device. That device joins the room’s group as a new ML-DSA-87 key that every member sees; it can’t sign as an existing device. If your own backend signs people in, how it does so is your choice.
What it costs
Section titled “What it costs”An ML-DSA-87 signature is 4,627 bytes and a public key is 2,592 bytes, against 64 and 32 bytes for Ed25519. The cost is paid once per handshake or proof, not per media frame, so it adds nothing to the media itself. Joining and key changes carry larger messages; streaming does not.
Related
Section titled “Related”- End-to-end encryption: the encryption layers in a room
- moqom-token and moqom-proof: the token and proof formats
- moqom-mls: room key agreement
- Threat model & controls