Authenticity, not just extraction.
Layer tamper detection, fraud signals, and chip-level checks on top of OnPoint OCR — without changing your integration. One response, one decision, one auditable trail.
Six categories of fraud, all in one call.
Verification ships as an additive flag on the extraction request — same auth, same JSON envelope, deeper signal.
Visual tamper detection
Pixel-level analysis of fonts, kerning, MRZ font widths, and substitutions. Flags photo-edits, recut bio pages, and bad copy-paste jobs.
Cryptographic checks
Where present (eMRTDs, NFC chips), we surface SOD validation, certificate chain status, and expected-vs-observed BAC keys.
Liveness adapters
Plug your own selfie / liveness vendor in, or use ours. We return a single combined risk score so policy lives in one place.
Signal you can act on
Per-check pass/fail, not a black-box score. Drives auto-approve, soft-decline, and refer-for-review without opaque ML excuses.
Fraud pattern learning
We feed back tampering patterns we see across customers (anonymized) so every tenant benefits from the latest fraud rings.
Same retention contract
Verification will run under the same rules as extraction: the image is processed in memory and never stored, and the structured result is kept for up to 30 days, then purged. Nothing separate, nothing longer.
A flat list of checks, not a magic score.
You decide which checks block, which warn, and which go to manual review. We just give you the truth.
- mrz_checksumPass / Fail / N/A
- mrz_font_matchPass / Fail / N/A
- photo_compression_consistentPass / Fail / N/A
- ghost_image_alignmentPass / Fail / N/A
- bio_page_substitutionPass / Fail / N/A
- issuing_state_template_matchPass / Fail / N/A
- doc_validity_windowPass / Fail / N/A
- stoplist_matchPass / Fail / N/A
ICAO 9303 · GDPR · SOC 2 · ISO 27001
The same compliance footprint as the extraction API. Sub-processor list, DPA, and security questionnaire on request.
Stop chasing fraud after the fact.
See what verification flags inside your existing OCR call.