What APX-DFA actually checks

APX-DFA returns a Pass, Warning or Fail determination on a scanned identity document. The question we are asked most often is what sits behind that answer.

The short version: eleven checks run against every document. They look at different things (the ink, the data, the layout, the logic of the dates), and a document can fail any one of them for a different reason. Together they cover the ways a document is usually wrong: it was never genuine, it was genuine and has been altered, it is a photograph of a screen, or it is real and simply out of date.

Capability follows the licence, not the device. The checks below are APX-Verify software results and require the corresponding APX-Verify licence.

1. Forensic data-driven validation

Multi-spectrum forensic analysis detects counterfeit patterns and synthetic IDs.

2. Multi-light (UV-IR) OCR security validation

Reads hidden text and security fields under multiple light conditions, for additional authenticity checks. Many security features are designed to be invisible under ordinary light and obvious under ultraviolet or infrared, so the document is read more than once, under more than one light.

3. 2D barcode (PDF-417) format authentication

Authenticates and cross-checks driver's licence barcode data against the printed fields. Supports barcode size, width, height, location, image analysis, and MRZ validation.

4. Document template detection

Identifies unknown or new ID formats by analysing layout and field patterns. Supports template and format validation. Issuing authorities change their designs; a document we have not seen before is not, by itself, a document that is wrong.

5. Cross-source data consistency

Compares key data fields extracted from multiple sources (OCR, barcode and MRZ) to verify that the values are identical. Detects tampering, reprints or mismatched templates by flagging differences in critical fields such as date of birth, document number and expiry date.

This is the check that catches the most common kind of alteration. A document carries the same facts in two or three places, encoded differently. Changing one of them convincingly is easy; changing all of them consistently is hard.

6. Field-level validation

Performs logical and format-based checks on each extracted field. Ensures all dates exist on the calendar, follow chronological order (issue precedes expiry), and that data conforms to valid formats: numeric, alphanumeric, ISO date. Flags incomplete or improperly formatted fields.

7. Geographic and jurisdiction validation

Validates geographic attributes (city, state, postal code, country) against official jurisdiction datasets. Confirms that the postal code and city combination is legally recognised and consistent with the issuing authority.

8. Calculated chronological metrics

Computes time-based analytics from the date fields, including current age, age at issue, validity duration and days since expiry. Supports eligibility checks and compliance logic.

9. Document colour and visual classification

Analyses the overall colour characteristics of the scanned image to ensure authenticity. Detects greyscale scans, altered overlays or low-quality reproductions inconsistent with official document standards.

10. Document status

Determines whether the document is valid, expired, or contains inconsistent or missing data. Indicates Expired when the expiry date has passed or when core field validations fail.

11. Jurisdiction and address validation

Confirms whether the city, postal code and jurisdiction align. Identifies address structure integrity and proper jurisdiction alignment.

Why a Warning is not a Fail

A Warning means something did not line up, not that the document is fraudulent. A genuine document photographed badly, an unusual but legitimate new template, or a field that did not read cleanly can all produce one. The result is designed to tell an operator where to look, not to make the decision for them.