SHA-256 Cryptanalytic Experiments — 60 Python Scripts (Noctarion 2026)

v1.1.0 (2026-04-28) — Reproduction script and corrections

Following critical review (BattleDog, Bitcointalk forum, 2026-04-27),
this version adds reproduce.py — a single self-contained reproduction
script with fixed seeds, pinned construction, and pre-registered claims.

Reproduction results at N=50,000:
  Finding #1 (cross-hash carry → LZ): r = +0.0018, |z| = 0.39σ — DOES
    NOT REPRODUCE. Original 9.56σ → 6.5σ → 0.39σ progression confirms
    the signal was sample-specific. Conceded: Finding #1 is dead.
  Finding #3 (round 8 insulator): largest avalanche jump locates at
    round 5 (Δ=0.125) in the pinned 72-byte / nonce=W[1] construction,
    not round 8 as claimed. Qualitative claim (sharp transition)
    survives; specific round number does not.
  Finding #5 (rotations dominate mixing): reproduces strongly. Ratio
    rot/Ch ≈ rot/Maj ≈ 4.78×10^8. Robust.

Net: 1 of 3 surviving findings is robust under fresh-seed reproduction.
Paper revision pending on IACR ePrint.

----

v1.0.0 (2026-04-26) — Original release

60 Python experiment scripts for the paper "Computational Irreducibility of SHA-256: 60-Experiment Cryptanalytic Study Across 19 Mathematical Frameworks" (IACR ePrint 2026/109079).

Each script is a self-contained cryptanalytic experiment on SHA-256 testing for statistical structure across 19 mathematical frameworks. Result: 0 exploitable signals.

Requirements: pip install numpy scipy z3-solver scikit-learn torch
Python 3.9+

Note (2026-04-27): The associated paper (IACR ePrint 2026/109079) 
has been revised — Finding #1 corrected to cross-hash carry 
anti-correlation r=−0.029, 6.5σ (internal state). The experiment 
scripts (exp12_double_hash_carries.py, exp62_strongest_signal.py) 
are unchanged and reproduce the corrected values.

Contact: Noctarion3@proton.me