Constant-Time Post-Quantum Solinas Cryptography and Incrementally Verifiable Silicon
Modern embedded, browser-hosted, and deep-space autonomy systems face six coupled cryptographic and verification crises: secret-dependent branches and lookup tables that leak Solinas field elements under differential power or timing analysis; unbounded recursive proof sizes when chaining astrodynamical, epigraphic, or automaton execution traces; naive 64-bit single-challenge Schwartz-Zippel soundness falling well below the 128-bit quantum-collision security floor; quadratic bloat when mapping 23-bit NIST FIPS-204 (ML-DSA-65) and 12-bit NIST FIPS-203 (ML-KEM-768) Number-Theoretic Transform (NTT) coefficients into 64-bit fields; IEEE-754 floating-point -0.0, subnormal Flush-To-Zero (FTZ), NaN, and ±∞ serialization equivocations across WebGPU/WASM host bridges; and multi-application domain-tag collisions and mathematical drift across heterogeneous software stacks. Building upon the 4/4 Quad-ratified release of Applied Purist Vol. V (10.5281/zenodo.24305230, Companion SW 10.5281/zenodo.24367776) and synchronizing the four Tier-1 Sovereign Quad-Directorates Dilaton Studio Core (137 Suites / 947 Tests GREEN) and Relativistic PNT Engine Lead (103 Suites/739 Tests [96/707] GREEN) as Co-Leads, alongside TriBody Workstation (83 Suites / 723 Tests GREEN) and Nath Darsana Core (67 Suites / 509 Tests GREEN) as Peer Auditors together with the three Tier-2 Auxiliary Triad nodes (ActOne Studio V1.0, 28 Suites / 152 Tests; Ancient Runes V1.0, 10 Suites / 109 Tests; Sub Translate V1.0, 12 Suites / 12 Tests; ecosystem total 440 Suites / 3,191 verified tests GREEN), this sixth applied volume establishes an OPSEC-hardened, mathematically unified, constant-time cryptographic and incrementally verifiable computation (IVC) architecture reserved under Preprint DOI 10.5281/zenodo.24392562 (with its dedicated Vol. VI Companion Software Suite DOI held strictly [Pending Post-Preprint Draft & 4/4 Archive Release]): (i) Dual Solinas Fields F_{2⁶⁴−59}/F_{2¹⁰⁶−117}, Canonical Quartic Tower F_{p₆₄}⁴, and 3-Pass Branchless Reduction: Proving primality and quadratic non-residue criteria (p₆₄ = 2⁶⁴ − 59 ≡ 5 (mod 8), (2/p₆₄) = −1; p₁₀₆ = 2¹⁰⁶ − 117 = 81,129,638,414,606,681,695,789,005,143,947 ≡ 3 (mod 4), (−1/p₁₀₆) = −1), we lock the canonical quartic tower F_{p₆₄²} = F_{p₆₄}[u]/(u² − 2) → F_{p₆₄⁴} = F_{p₆₄²}[v]/(v² − u) ≅ F_{p₆₄}[x]/(x⁴ − 2) (9 F_{p₆₄} Karatsuba multiplications, Frobenius ζ = 2^{(p₆₄−1)/4} Itoh-Tsujii inversion) and F_{p₁₀₆²} = F_{p₁₀₆}[u]/(u² + 1). We reconcile 2-pass carry overshoots (≤ 2⁶⁴ + 3421 and ≤ 2¹⁰⁶ + 13571) versus strict 3-pass bounds ([0, 2⁶⁴ − 1] and < 2p₁₀₆) prior to unsigned borrow-safe masked canonical subtraction across WGSL vec4<u32> and WASM i64x2 (0 bit defects across 128 boundary vectors). (ii) 106-Bit Knuth-Dekker D₁₀₆ → F_{2¹⁰⁶−117} Bridge and Bitwise rotl_64 Conservation: Injectively quantizing relativistic ephemerides (ε_dd = 2⁻¹⁰⁶) via 53+53-bit fixed-point scaling (|Q| ≤ 2¹⁰⁵ − 59), −0.0 → +0.0 normalization, and fail-closed NaN/±∞ traps maps D₁₀₆ unambiguously into F_{p₁₀₆} (reconciling dimensionless L_TCL = 6.4839 × 10⁻¹⁰ → 0x591d32 at S = 2⁵³ and 0xb23a63d97cf65800000 at S = 2¹⁰⁶ distinctly from operational lunar clock slope +56.020896 µs/day → 0x700ab2e1693c041 at S = 2⁵³ and 0x1565c2d2780828edc0298a9fbc at S = 2¹⁰⁶), while circular 64-bit bit-rotation preserves popcnt64(rotl64(x, k)) = popcnt64(x) and exact bijective inversion across all 64/64 shifts. (iii) Black-Box Oracle Profile of TMD-HASH V12.0, Disjoint Domain Registry, Canonical Test Vectors, and Open Companion-Software Boundary: Treating TMD-HASH V12.0 strictly as an opaque black-box Triple-Anchor permutation oracle (permanently preprint-only with zero TMD-Hash software DOI ever minted, and proprietary internal permutation equations, round/Feistel/exponent schedules, MDS/expander matrices, capacity slot layouts, and NUMS scripts withheld per isProprietaryLatexProtected: true), we retain strictly system-level cleared oracle bounds (B ≥ 8, A_S ≥ 80, MEDP < 2⁻⁴⁹³⁶, MELP ≤ 2⁻³⁵⁰⁰, SAC 50.04% at χ² p = 0.982, Groth16 320 [85.125× SHA-256] / Plonk 160 / STARK 384, ~3.2 cycles/byte AVX2), freeze a unified Two-Tier Wire Grammar (CH_A|CH_B|CH_C plus AUX-VOL6-* adapters), a disjoint Domain Tag Registry, RFC-8785 / length-delimited payload canonicalization, UTF-8 vs. seeded Table 4 vector provenance, and an explicit software boundary ensuring Vol. VI's future Companion Software DOI ships solely open Solinas/IVC/NTT/ABI verification harnesses without proprietary TMD-Hash source code. (iv) O(1) Relaxed-R1CS Folding (tmd12-ivc65 to Nova IVC-107), F_{p₆₄}⁴ Soundness (2⁻²⁴⁶.⁷³), and FIPS-203/204 46-Bit Packing: Relaxed-R1CS folding with exact cross-term T achieves zero residual slack and 148.75× gate compression (256 constraints/step vs. 38,080 unrolled, 0.42 ms verifier). For effective degree D_eff = 619, extension challenges raise Schwartz-Zippel soundness from single-field 2⁻⁵⁴.⁷³ to 2⁻²⁰².⁷³ (F_{p₁₀₆}²) and 2⁻²⁴⁶.⁷³ (F_{p₆₄}⁴), strictly dominating the 2⁻¹²⁸ security floor. Bijective 46-bit packing w_{a,j} = c_{a,2j} + 2²³ c_{a,2j+1} bounds w_max = 70,300,033,081,344 < 2⁴⁶ < p₆₄ < p₁₀₆ (266/266 vectors, 64.2 µs ≤ 65.0 µs verify, 112 B ECDSA P-256 telemetry chain). (v) Multi-Target Constant-Time Silicon, TVLA (|t| < 4.5), SP 800-88 Wipe, MPS/RQS Audit, and SolinasIvcState64 64B ABI: Separating hardware-compiled WGSL select(), WASM SIMD i64x2 (16–64 static pages), MISRA-C99 __uint128_t, Rust #![no_std], and SystemVerilog RTL (TVLA |t| < 4.5, 10k-node WebGPU batch 1.42 ms < 1.50 ms) from non-CT TypeScript BigInt reference models, we certify NIST SP 800-88 buffer zeroization (0 residual bits), N = 32 MPS / 147,456-weight RQS Merkle authentication (384 B path, 0.19 ms), and Canonical Listing 7.1 SolinasIvcState64 (64 B, read-only bitcast<u32> & 0x3fffffffu across 6/6 Morton-30 keys, CCSDS OEM/OPM, 0 B GC). (vi) Tier-2 Auxiliary Profiles (AUX-VOL6-*) & Joint 6-Gate Certification (IVV-APP6-01..06): Within their Purist Vol. V (10.5281/zenodo.23762332) + TMD-HASH V12.0 horizon, SubTranslate (|Δp_i| ≤ 2/255, −42.11 dB, disjoint TMD12_QR_PAY envelope, 0 B LL(1) GC), ActOne (injective Artin B_n pure-braid Lk = 0 + Fountain AST uniform ~14-R1CS step, migrating to F_{p₆₄}⁴), and Ancient Runes (1,551 B zero-import WASM DSP, 0.0019–0.0680 cents, 1.16–1.23 ms/frame, V1 errata resolved) complete the unified six-gate matrix advancing Applied Purist IV&V from 65 → 71 gates.
Authors
- AryaArunachalaAnanda Ghulam-e-Shah-e-Unmani (ORCID: https://orcid.org/0009-0004-5288-3949)
- Aghora Abraham Global LLC
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24392561
- Primary Topic
- Cryptography and Residue Arithmetic
- Type
- preprint