Icarus Post-Quantum Crypto

Icarus is the post-quantum cryptography engine of Syn_OS. v9.0 implements all three NIST-standardised PQC algorithms and exposes them through a single Rust crate (synos-icarus) used by every component that signs, encrypts, or attests anything.

The name is the test: fly toward the sun if you must, but make sure your wings are made of something that doesn’t melt under quantum heat.

What’s implemented

Algorithm	Standard	Class	Use in Syn_OS
ML-KEM	FIPS 203	Key encapsulation	TLS 1.3 hybrid, ARCANUM mesh peer agreement, Sanctum federation tunnels
ML-DSA	FIPS 204	Module-lattice digital signature	ALFRED model provenance, lab integrity manifests, audit-trail signatures
SLH-DSA	FIPS 205	Stateless hash-based signature	Long-lived release signing where statefulness is unacceptable

Why three?

Different PQC families have different security assumptions. Syn_OS uses all three so that a future cryptanalytic breakthrough in one family does not collapse the entire trust chain.

ML-KEM — relies on the hardness of Module Learning With Errors (MLWE). Fast, small ciphertexts. Default for transport security.
ML-DSA — also lattice-based. Larger signatures than ECDSA but acceptably small. Default for high-frequency signing (audit trail entries, log signatures, mesh attestation).
SLH-DSA — purely hash-based. Conservative, no lattice assumptions, but with much larger signatures (~30 KB) and slower signing. Used for release artefacts and other long-lived signatures where the future risk of lattice cryptanalysis is unacceptable.

Where it’s wired in

Subsystem	Algorithm	Notes
TLS terminators (Sanctum federation, ARCANUM mesh)	ML-KEM hybrid with X25519	rustls-pqc fork, hybrid key share
Curtain v3 capability tokens	ed25519 today, ML-DSA migration prepared	Token format already carries an algorithm OID for clean rotation
Audit trail (`synos-audit-trail`)	ML-DSA per entry, SLH-DSA on epoch roots	HMAC-SHA256 chained, signed roots
Build attestation (v48 Forge)	SLH-DSA on release manifests	Signed alongside cosign / Sigstore
ALFRED model provenance	ML-DSA on ONNX / Ollama model fingerprints	v42 Carcossa Burning
Lab integrity manifests	ML-DSA on SHA-256 manifest roots	Verified by `lab-integrity` xtask
Fragment Field IDS reports	ML-DSA per report	Mesh-distributed, must verify across peers

Performance

Measured on Haswell-era i7-4790 (the THEDARKNESS admin node baseline):

Operation	ML-KEM	ML-DSA	SLH-DSA
Keygen	47 µs	61 µs	6.8 ms
Encapsulate / Sign	62 µs	88 µs	140 ms
Decapsulate / Verify	78 µs	175 µs	12 ms
Public key size	800 B	1312 B	32 B
Ciphertext / Signature size	768 B	2420 B	~30 KB

For high-frequency signing (audit trail entries) ML-DSA is the hot path. SLH-DSA is reserved for slow paths where 140 ms per signature is acceptable in exchange for the hash-based security floor.

Hybrid mode

For transport security Icarus runs hybrid — combining classical (X25519 / Ed25519) with the post-quantum algorithm — so that if either primitive falls, the connection is still secure under the other. This matches NIST and IETF guidance for transition deployments.

use synos_icarus::tls::HybridSuite;

let suite = HybridSuite::X25519MLKEM768;       // transport
let signer = HybridSigner::Ed25519MLDSA65;     // signing

The synos-icarus crate exposes a stable Rust API that the rest of the workspace uses; consumers don’t need to know whether the underlying primitive is pure-PQC, hybrid, or classical. The choice is made by configuration policy.

What is not implemented

CRYSTALS-Kyber-512 / Dilithium2 legacy parameter sets — superseded by FIPS 203 / 204; not shipped
Falcon / FALCON-512 — NIST round-3 alternate; available as a feature-flagged extension crate but not default
Stateful hash-based signatures (LMS / XMSS) — explicitly avoided. Statefulness is operationally hostile.

Quantum-safe today vs quantum-safe tomorrow

Icarus is quantum-safe today in the sense that every signing and key-agreement primitive that defends Syn_OS infrastructure is either FIPS-203/204/205 or a hybrid that includes one of those. There is no classical-only path in production Sanctum federation traffic, audit trail, or release attestation as of v60.

The migration of Curtain v3 capability tokens from ed25519 to ML-DSA is the last remaining transition; it is feature-gated behind a config flag pending counsel review of token format compatibility with downstream MSA terms.

Custom Kernel → — synos-icarus is also linked into kernel modules for module-signing
Forge → — SLH-DSA on release artefacts
Curtain → — capability token signing