Confidential — Pre-disclosure Internal prototype. No public discussion, demo, or embedding outside NDA until USPTO provisional is filed.
ARIA-AUT · Cross-cutting Modifier
AIe BinômeCooperative Cross-Verification
Two or more independent verifiers · independent re-derivation · consensus-gated authentication · AI-forgery resistant
VOXEL-DOC-2026-001 · Modifier — Phase 5
Carries from · MP19 P88 (Binôme parent)
Applies across · Embodiments A · B · C · D · F
Locale · Geneva · 2026-04-26
Verifiers
3
Independent computational entities
Consensus mode
2-of-2
Unanimous strict
Independence axis
Architectural
Different model classes
Adversarial-AI cost
×N harder
Must fool all consensus members
Counterparty Enrollment
SETUP

Each enrollment generates a 256-bit secret as an HSM stand-in.

Document Issuance
ISSUE

Document hash is bound to canonical content. Each verifier independently re-derives expected addresses + values.

Consensus Configuration
PROTOCOL

Stricter modes reject more aggressively; more lenient modes tolerate transient verifier outages.

Verification
VERIFY

All active verifiers re-derive the expected addresses & values independently from the counterparty secret & doc hash.

Targeted = adversarial AI optimised to fool one specific verifier. AI cross-fool = attempts to fool all verifiers (much harder). Tamper = changes document content.

VERIFIER A
Pipeline-CV
·
awaiting
Classical computer-vision pipeline · deterministic feature extraction · rule-based matching
Match score
score —
accept ≥ 85%
VERIFIER B
Neural-Transformer
·
awaiting
Attention-based feature classifier · learned decision boundary · pattern-driven
Match score
score —
accept ≥ 90%
VERIFIER C
Signal-Processing
·
awaiting
Frequency-domain analyser · wavelet decomposition · spectral fingerprint
Match score
score —
accept ≥ 80%
·
Binôme consensus
awaiting
Issue a document and run cooperative verification.
0 / 0 vote
accept · 0
reject · 0