MLKEMCipher: Konstanter IND-CCA2-Kern, vektorisierte NTT & Barrett-Reduktion

MLKEMCipher liefert abhängigkeitsfreie, zeitkonstante Referenzimplementierungen und Assembler-optimierte Kernel für die ML-KEM-Polynomarithmetik über dem Ring R_q mit q = 3329 und n = 256 mit nachweisbarer IND-CCA2-Sicherheit.

MLKEMCipher Lattice Arithmetic Pipeline & Vector Processing Topology

Ring R_q polynomial operations, Number Theoretic Transform (NTT), and Montgomery reduction

Central System: Lattice Vector Engine - AVX-512 / NEON Core

Cryptographic Nodes

  • Forward NTT Unit
  • Pointwise Multiplier
  • Inverse NTT (INTT)
  • Centered Binomial Sampler
  • FO Transform Engine

Architectural Layers & Protocols

  • Polynomial Vector Ring R_q (R_q = Z_q[X]/(X^256 + 1), q=3329): Defines modular lattice algebra with k=2 (512), k=3 (768), and k=4 (1024) matrix ranks.
  • Number Theoretic Transform (Radix-2 Cooley-Tukey NTT with Twiddle Precomputation): Reduces polynomial multiplication complexity from O(n^2) to O(n log n).
  • Fast Modular Reduction (Montgomery Reduction (R = 2^16) & Signed Barrett Reduction): Executes coefficient division and modulo 3329 without hardware divide instructions.
  • Fujisaki-Okamoto Transform (IND-CPA to IND-CCA2 Conversion with Implicit Rejection): Re-encrypts decrypted plaintext to verify ciphertext integrity before key extraction.

Key Metrics: q = 3329 (Prime Modulus) | n = 256 (Polynomial Degree) | O(n log n) (Multiplication Complexity) | 0 Branch (Data-Dependent Branches)

Konstantzeitige IND-CCA2-Chosen-Ciphertext-Sicherheit, NTT-Polynomvektorarithmetik und Montgomery/Barrett-Modularreduktion

Autoritative technische Referenz fuer constant-time ind-cca2 chosen-ciphertext security, ntt polynomial vector arithmetic & montgomery/barrett modular reduction gemaess NIST FIPS 203/204/205, gitterbasierter Kryptographie und Seitenkanal-Haertung.

Core Engineering Areas

Technical Articles

Related Cluster Architecture Specifications