DocHorizon Ideas Portal
Add a new idea Subscribe
Categories New feature
Created by Thijs Horstman
Created on Sep 18, 2025

RELATED IDEAS

Fraud Detection new features

See related ideas for new feature overview

  • Attach files
  • Guest
    Feb 5, 2026

    • Explainable AI: Providing natural language descriptions of why a document was flagged (e.g., "The font on the date field does not match the rest of the document").

    Reply
  • Guest
    Feb 5, 2026

    • Template Matching: We could maintain a "blocklist" database of known internet receipt/invoice maker templates.

    • Screenshot Detection: We could analyze EXIF data, aspect ratios (screen dimensions), and resolution (72 DPI vs print DPI) to flag screenshots.

    Reply
1 MERGED

Fraud: own AI model for generative AI detection

Merged
Next to the C2PA metadata check, we need a new model to check if a document was created with generative AI tools.
Thijs Horstman 10 months ago in New feature 0 Shipped
1 MERGED

Fraud: external (cross)checks

Merged
Implement external APIs to validate Google Maps locations, check if someone is on a terrorism or PEP list
Thijs Horstman 10 months ago in New feature 0 Future consideration
1 MERGED

Fraud detection based on expected pricing range

Merged
Ask an AI model if the extracted line item price is within the expected price range for this product/service. If it's not, this could be a fraud indicator. For example: iphone repair for €2000. A service like that would normally cost around €50.
Guest 10 months ago in New feature 0
1 MERGED

Fraud: send in userdata or context

Merged
Alongside the document/image, send in additional information with the request to validate against. For example, send a list of allowed phone brands with which the image has to be taken, or send in countries to be validated with geodata from the EXIF.
Thijs Horstman 10 months ago in New feature 0 Future consideration