AI Document Verification

MRZ Code Reader

MRZ or Machine Readable Zone is a particular area in an identity document (Passports and ID cards) that encloses the document holder’s personal data. Our AI Document Verification solution can easily decodify it.

• Name
• Surname
• Date of Birth
• Country
• Document number
• Nationality code

Data Matching

Once information is extracted from visual and MRZ Sensity AI Document Verification will automatically match those data to ensure the document hasn’t been altered.

Data format validation

In the case of a manipulated document, it may happen fraudsters are not accurate in change data. Each data on the ID document has a precise format compliant with the national regulation (caps, font type, number of characters etc). With Sensity’s AI document verification you can detect invalid data format on thousands of documents worldwide.

Recency Check

Sensity AI document verification returns and instant rejection if the document has expired or it doesn’t fit with the company expectations.

Authenticity Checks

This is one of the most powerful features in our AI document verification suite. For most of the ID formats we are able to extract one of the most significative parts and through  computer vision algorithms match them with the same part extraded by the original format issued by  the government. If the algorithms give us a high confidence score, we consider that check passed. If the compared parts are not considered identical. The document will be rejected.

Ghost portrait analysis

If the document format presents ghost portrait we can send two warnings in case the ghost portrait is missing (in contradiction with the national format that needs a ghost portrait) or the ghost portrait doesn’t match with the main picture, so it can be considered a spoofing attempt as you can see in the examples below.

Metadata analysis

We investigate the file looking for signals of a potential forged document. This suite include 8 different checks to improve your fraud detection capabilities.

Splicing Detection

This approach is entirely based on a deep neural networks able to analyze the document at pixel level, detecting manipulated fields.