Solutions · HSM & Hardware · For vendors and enterprise buyers

One engine for every prime. Not four patched together.

Today's HSM vendors don't ship a cryptographic engine, they ship a patchwork: one code path for RSA primes, another for NTT-friendly FHE parameters, another for zk-SNARK moduli, and soon a fourth for post-quantum migration. Four code paths. Four audit trails. Four optimization targets. Four places the performance ceiling gets hit.

Ethoryx ships the Unified Prime Engine: a single C core that serves every cryptographic primitive the HSM needs. Integrated through standard PKCS #11 C_GenerateKey(). Zero application changes.

1. The Patchwork Problem 2. The Unified Engine 3. Zero-Migration Integration
Pillar 1 · The Patchwork Problem

HSM vendors don't have a single brain.

The cryptographic engine inside a top-tier HSM today is a fragmented architecture: separate code paths for each primitive, each tuned independently, each auditable only in isolation. What looks from the outside like "a cryptographic module" is internally four silos that happen to share a chassis.

Current architecture · typical HSM vendor
Application  →  C_GenerateKey(mechanism)
Silo 1
RSA Path
Miller-Rabin · OpenSSL BN · hand-tuned for 2048/4096
Silo 2
NTT / FHE Path
Lattice parameters · NTT-friendly prime search · SEAL bindings
Silo 3
zk-SNARK Path
BN254 / BLS12-381 · curve moduli · proving-system specific
Silo 4
PQC Path (forthcoming)
Kyber / Dilithium moduli · lattice sampling · NIST parameter sets

Performance ceiling is the slowest silo

When a customer enables FHE operations alongside RSA, their throughput collapses to the weakest path. The RSA team can't optimize the FHE team's code. Each silo's ceiling becomes the product's ceiling.

Four audits, not one

FIPS 140-3 and Common Criteria certify code paths. Four silos mean four certifications, four re-certifications on every change, and four places a CVE can hide. Cost and time scale with silo count, not with capability.

PQC migration hits a wall

As NIST PQC becomes mandatory through 2027, vendors must add a fifth silo. Each new silo is a new multi-quarter project. The patchwork cannot absorb post-quantum gracefully, it must be extended, audited, and re-sold.

Pillar 2 · The Unified Engine

One core. Every prime.

Ethoryx consolidates prime-generation into a single C core with a single optimization target. RSA primes, NTT-friendly primes for lattice FHE, curve moduli, and post-quantum parameters all flow through the same engine. One code path. One performance story. One audit surface.

Ethoryx architecture · one engine
RSA
C_GenerateKey · CKM_RSA_PKCS_KEY_PAIR_GEN
FHE / NTT
CKM_GENERIC_SECRET_KEY_GEN (custom)
zk / Curves
Vendor mechanism
PQC
NIST draft mechanisms
Ethoryx Unified Prime Engine
One C/GMP core · thread-safe · constant-time candidate selection · theorem-guided sieve

Single optimization target

When we improve the engine, every primitive gets faster simultaneously. The RSA path and the FHE path share the same sieve, the same candidate selection, the same GMP calls. One optimization surface, not four.

One audit, one certification

FIPS 140-3 and Common Criteria evaluate one C core against one threat model. New primitives inherit the certification instead of requiring new ones. Audit cost becomes constant, not linear in silo count.

PQC-ready by construction

The engine treats prime generation as a single mathematical problem; finding integers that satisfy specified constraints. Adding a post-quantum primitive means adding constraints, not adding a silo. NIST PQC migration becomes a configuration change.

Pillar 3 · Zero-Migration Integration

Standard PKCS #11. No application changes.

Enterprise applications already call C_GenerateKey(). They don't need to know which engine is behind the PKCS #11 interface. Ethoryx slots in underneath application code stays identical, the engine changes, performance and audit properties improve. Drop-in replacement, not migration project.

Customer application · unchanged
// Existing enterprise code not modified
CK_MECHANISM mech = { CKM_RSA_PKCS_KEY_PAIR_GEN, NULL_PTR, 0 };
CK_OBJECT_HANDLE pubKey, privKey;

C_GenerateKey(session, &mech, template, attrCount, &pubKey, &privKey);
// ↑ This call used to take ~280ms on 4096-bit RSA.
//   With Ethoryx engine behind PKCS #11, same call, faster, auditable, PQC-ready.
Step 1

Drop-in replace

Ethoryx engine integrates behind the HSM vendor's PKCS #11 layer. Application-visible API is unchanged.

Step 2

Route by mechanism

CKM_RSA_PKCS_KEY_PAIR_GEN, NTT primes, curve moduli of all routed to the same engine with different constraints.

Step 3

Inherit certification

Engine carries its own FIPS / CC evidence. HSM-level certification incorporates Ethoryx artifacts rather than re-auditing from scratch.

Step 4

Extend via config

Future primitives (PQC, new FHE schemes, emerging zk systems) added as engine configurations, no silo expansion, no new audit surface.

Reference engineering figures

Measured on the Ethoryx reference engine under controlled evaluation. Customer-specific results vary by workload and hardware; request a technical evaluation for figures on your target configuration.

Engine core
1 C/GMP
One unified core serving every prime-generation primitive. No per-primitive silos.
Candidate selection
Constant-time
Selection of prime candidates runs in constant time, eliminating timing side channels common in OpenSSL-based HSM paths.
Thread-safety
Verified
Engine is reentrant. Multiple PKCS #11 sessions can request generation simultaneously without serialization.
PQC migration
Config, not rebuild
Adding post-quantum primitives is a constraint change in the engine, not a new silo. Vendor roadmap unblocks.

Stop maintaining a patchwork. Ship one engine.

Ethoryx Research partners with HSM vendors and enterprise buyers on technical evaluation, integration pilots, and long-term engineering arrangements. If your roadmap includes PQC migration, FHE at scale, or simply reducing the audit surface of your cryptographic module, this is the conversation to have in 2026.