Aller au contenu principal
From Shannon to silicon

Cryptographic
fundamentals

27 foundational papers, 9 layers of proofs, 3 independent mathematical families.

Cryptography is not a product. It is a trust chain that starts with a theorem and ends in silicon.

Formal construction

Our Construction Π

Multi-KEM Combiner with PSK — integrated architecture with code-based component activable via signed firmware update upon NIST standard publication.

Construction Π — Multi-KEM Combiner with PSK
ss = H(ss1 ∥ ss2 ∥ ss3 ∥ ψ ∥ H(ct1) ∥ H(ct2) ∥ H(ct3))
K1, K2, K3 — three IND-CCA2 KEMs from distinct mathematical families
ψ — 256-bit pre-shared key (unconditional security)
H — ciphertext-binding extractor (CR + PRF)
cti — per-component ciphertext binding
Theorem: Π is IND-CCA2 secure if any single component holds.
AdvΠ ≤ min(AdvK1, AdvK2, AdvK3, 2−256) + negl(λ)
Total break requires: 3 independent mathematical breakthroughs + 1 physical theft
Trust chain

9 layers of proofs

From information theory to international certification. Each layer builds upon the previous one.

0
Layer 0

Theoretical foundations

Shannon 1949 — Information theory

"A perfect secret exists." Our PSK is one.

1
Layer 1

Hard problems

Ajtai 1996 · Regev 2005 · Peikert 2009 · Berlekamp-McEliece-van Tilborg 1978

Three families of mathematical problems studied for 30 to 78 years. No polynomial algorithm found. The best attempt (Chen 2024) was retracted.

2
Layer 2

Cryptographic constructions

Fujisaki-Okamoto 1999 · Jiang et al. 2021 · Barthe et al. 2023

Proven transformation: a CPA encryption becomes a CCA2 KEM. Machine-verified proof by INRIA (EasyCrypt).

3
Layer 3

Combiner

Giacon-Heuer-Poettering 2018 (PKC)

Our central theorem: combining N KEMs = security of the strongest. Ciphertext-binding: each component is cryptographically linked.

4
Layer 4

Independence

Cramer-Ducas 2016 · Hou-Jiang 2026 · MATZOV 2022

Proof that our 3 families do not fall together. Real attacks (Hou-Jiang: 7-13 bits) that justify our diversification.

5
Layer 5

Protocol

Dowling-Paterson 2018 · Noise Framework · Rosenpass 2023

Formally proven handshake in the eCK model. 100% FIPS primitives. Zero unapproved algorithms.

6
Layer 6

Implementation

Huang et al. 2022 (Plantard) · KyberSlash 2023 · dudect 2017

Assembly code verified bit by bit. Zero divisions in the binary. Proven constant-time — no timing leaks.

7
Layer 7

Formal verification

EasyCrypt (INRIA) · F*/hax (MSR+INRIA) · Vale · CompCert (INRIA)

Each proof is itself proven by a machine. Our verification chain relies on tools from European and international academic research.

8
Layer 8

Concrete estimates

Lattice Estimator (Albrecht 2024) · Both-May 2018 · Chailloux-Loyer 2021

Every security claim has a number. Every number has a calculation. Every calculation is reproducible.

9
Layer 9

Certification

NIST FIPS 203/204/207 · ANSSI PA-2024-007 · NSA CNSA 2.0

A single design certifiable across 3 continents.

Methodology

Analyse. Understand. Verify. Certify.

Analyse

Every publication evaluated by our tools. Lattice estimator, ISD calculator, concrete estimates.

Understand

Every attack reproduced and numerically verified. 560 tests, 0 failures.

Implement

Bare-metal assembly. AMD FPGA. Zero divisions. Proven constant-time.

Certify

ANSSI CSPN (Q3 2026). FIPS 140-3 Ready. CNSA 2.0. SOG-IS + CCRA (31 countries). One design, the entire world.

Want the details?

The complete cryptographic justification dossier — formal proofs, concrete estimates, verification specs — is available under NDA for certified evaluators and qualified clients.