The 34-Point Verification Scanning Protocol
Every document submitted to Testiy — from job offer letters to work permits to employment contracts — runs through a three-stage pipeline: programmatic preflight, parallel intelligence, and dual-model AI verification consensus. Deep binary analysis and cross-document validation catch forgeries a manual eyes-only review cannot.
Stage 1
Programmatic Preflight
Stage 2
Parallel Intelligence
Stage 3
AI Verification Consensus
34
Verification Checks
3
Pipeline Stages
Dual
Model Consensus
24h
Auto File Deletion
Three Stages. Thirty-Four Checks.
The live Testiy engine expanded past the original 18 verification lenses into a full 34-point pipeline — programmatic first, intelligence next, AI consensus last.
Programmatic Preflight
Byte-level analysis before any AI runs. File integrity, SHA-256 fingerprints, PDF revision history, design-tool traces, EXIF/encoder fingerprints, clone-stamp detection, print–rescan classification, and known fraud-template matching — all deterministic and auditable.
Parallel Intelligence
Independent intelligence passes that share no prior conclusions. Cross-document field locks, letterhead consistency, employer registry lookups, sponsor/web presence signals, domain & mail-server traces, salary plausibility, and entity relationship mapping.
AI Verification Consensus
Dual-model AI consensus on pixel & typography verification, geometry, date/permit logic, official markings, identity consistency, legal text authenticity, country compliance, fraud patterns — then a weighted risk synthesis and final verdict.
Built for What Actually Gets Rejected
Flight itineraries and hotel bookings can be double-checked against a free embassy checklist in five minutes. Work permits, job offer letters, and European employment contracts cannot. Detecting a fabricated company registration number, an offer letter that doesn't match the destination country's statutory contract clauses, or a work permit category that doesn't legally exist requires expert-grade document analysis — which is exactly what our engine is built for.
That's why our scanning protocol is weighted toward the categories our applicants are rejected for most often: employment authenticity, statutory compliance, and employer legitimacy — not just formatting.
18 Core Verification Lenses
The classic analysis layers still power Stage 3 AI consensus — now wrapped inside the fuller 34-point pipeline with preflight and parallel intelligence.
Pixel Integrity
Analyses raw compression data of the image. When a region is copy-pasted or edited, block boundaries stop aligning with the rest of the image. We look for mismatched noise profiles, clone-stamp artefacts, and colour-grading signatures — manipulation invisible to the human eye.
Typography
Checks that every character comes from the same font, rendered the same way. Genuine government and corporate documents use a single consistent typeface. Forgeries assembled from multiple sources mix fonts with different stroke weights or kerning — including invisible Unicode substitution attacks.
Metadata & Origin
Reads every embedded data layer in the file to classify a document's true origin: a genuine scan, a born-digital official file, or a digital / reprinted forgery. Verifies the actual binary format matches its file extension.
Geometric Structure
Measures perspective, margin consistency, and the parallelism of ruling lines and text baselines. A genuinely scanned document has consistent geometric distortion; a composite shows subtle misalignments that don't match a single physical scan.
Date & Temporal Logic
Validates every date against country-specific rules. Work permits, job contracts, and visa categories each have legally defined validity periods — we flag future issue dates, impossible validity lengths, and overlaps that don't add up.
Official Markings
Examines stamps, visa labels, holograms, and embossed seals. Genuine stamps show consistent ink diffusion; digitally pasted stamps have perfectly uniform edges and no ink spread.
Identity & Biometric
Validates every identity field across the document, including MRZ check digits, visual-to-MRZ consistency, photo integrity, and cross-referenced dates of birth.
Editing Artefacts
Looks for traces left behind by image editing: fill-layer overlays, selection-feathering halos around pasted content, and layer-flattening signatures.
Third-Party Editor Fingerprint
Identifies whether a file was opened and saved in a consumer editing application — each leaves a distinct fingerprint in the file's underlying structure.
Text Authenticity
Audits legal language in job contracts and offer letters against jurisdiction-specific templates — regulatory references, generic filler language, and internal contradictions are all flagged.
Cross-Document Consistency
Reconciles every shared field across all submitted documents. Full name, employer name, salary figures, and reference numbers must chain together flawlessly across your entire dossier.
Security Features
Checks for physical security printing techniques used in genuine government and corporate documents — patterns that are extremely difficult to replicate digitally.
Cold-Start Reconstruction
Detects documents fabricated entirely from scratch — font authenticity scoring, letterhead fidelity, and missing legally required clauses all factor into this test.
Work-Permit Compliance
Validates that the claimed permit type actually exists in the destination country and that the stated salary meets the legal threshold for that permit category and occupation.
Company Intelligence
Extracts the employer's registration details and validates them against official national and EU business registries — confirming the company is real, active, and consistent with the role offered.
Print-Rescan Attack Detection
Detects the specific artefact pattern left when a digitally edited document is printed and rescanned in an attempt to launder digital manipulation traces.
Fraud Pattern Intelligence
A specialised module targeting high-volume fraud patterns seen in fabricated job offers and work-permit schemes — fake recruiter databases, illegal recruitment-fee clauses, and other known scam signatures.
Risk Summary
Synthesises scores from preceding tests into a single trust score and a clear verdict — approve, manual review, or reject — with a full breakdown of every flag raised.
From Upload to Verdict
Every document takes the same five-stage path through the engine — fully automated, fully auditable.
Secure Upload
Encrypted intake — files are quarantined per-session and never shared.
Programmatic Preflight
17 byte-level checks run before any model sees the file.
Parallel Intelligence
Registry, domain, and cross-document passes run in isolation.
AI Verification Consensus
Dual-model review with no shared bias before verdict.
Verdict & Report
A trust score plus a full breakdown of every flag raised.
Specialised Verification Modules
High-risk document elements — stamps, signatures, and employer identity — go through dedicated multi-layer verification on top of the core 34-point scan.
Stamp & Seal Verification
- Visual inspection for ink diffusion, pressure variation, and imperfect geometry that only a physical stamp produces — digital stamps have razor-sharp, uniform edges.
- Binary-level analysis of the underlying file bytes to detect the software signatures left behind by editing tools.
- Every stamp is cryptographically hashed and cross-referenced against previously flagged fraud cases.
Signature Verification
- Classifies signatures as hand-drawn or digitally applied — physical signatures show organic stroke-velocity variation that digital signatures lack.
- Error Level Analysis re-compresses the document and amplifies pixel differences, revealing pasted signature regions as bright hotspots.
- Cross-checks signatory identity against the rest of the document for consistency.
Company Intelligence
- Queries official registries across multiple countries to confirm a company exists, is active, and its registered address matches what's on the document.
- Checks whether the employer is actually authorised to sponsor work permits under the destination country's official sponsorship registers.
- Domain and email verification — flags newly registered domains, free-email HR contacts, and other employer red flags.
- Cross-references new cases against known fraud patterns to catch coordinated fraud rings spanning multiple applicants.
- Validates salary figures against country minimum wages and flags statistically improbable round-number salaries.
Multiple Independent Passes. No Shared Bias.
Every high-risk finding is verified through multiple independent analysis passes that never see each other's conclusions before reaching their own — so a single faulty assumption can't silently pass through the entire pipeline unchecked. Where two independent passes disagree, the case is automatically escalated to a human compliance specialist for manual review rather than being auto-approved.
The exact configuration of our scanning pipeline — including model selection, weighting, and escalation thresholds — is proprietary and not disclosed publicly.
Ready to Have Your Documents Scanned?
Work permits, job offer letters, and employment contracts get our deepest level of verification review — the categories most likely to be forged, and hardest to self-verify.