Photo Verification

This service is based on biometric face recognition technology – Microsoft Face API.

To perform the check, this service requires identification data (TIN, full name, date of birth) and the photo of the person.

The check is performed by making a comparison and determining the degree of similarity of the provided photo, with:

• the reference photo from the UBCH’s database of subjects of credit histories, according to the specified data,
• all the photos from the fraudster database,
• all the photos from the UBCH’s database of application photos (persons who applied for credits to the partners of the Bureau, without a reference photo for verification),
• all the photos from the open database of missing and wanted people of the Ministry of Internal Affairs.

Experiments using training sets, showed that:

• When comparing with the reference face photo (one-to-one), the sufficient similarity degree is 32%.
• When comparing with a set of face photos, from each of the databases (one-to-one), the sufficient similarity degree is 70%.

Therefore, the conclusion about successful photo verification is made, if:

• the degree of similarity with a person’s reference photo exceeds 32%,
  no matches with the similarity degree exceeding 70% were found in the fraudster database and the database of missing and wanted people,
• no matches with the similarity degree exceeding 70% were found in the database of application photos, with the identification different from the specified one.

In addition to the basic mode of the service, the following options can be used:

• Checking the provided photo, by comparing it with only the reference face photo from the organization’s own database (if one was submitted beforehand to the bureau);
• Search only in the photo-database of compromised persons (fraudsters, missing people, wanted people);
• Comparison of two incoming photos;
• Using the photo from the document as the provided photo;
• comparison of two persons on one photo, for example, the face in the passport and the face of the person holding it.

• Identity verification at first contact;
• Remote verification for online interaction;
• Additional verification for identity confirmation.

Photo-verification allows you to significantly reduce risks.
Hit rate of a photo availability in UBCH’s database is 70%,

• 0.82% of requests reveal mismatch with the reference face photo,
• 0.05% of requests reveal matches with the fraudster database,
• 0.04% of requests reveal matches with the database of wanted people,
• 0.01% of requests reveal matches with the database of missing people.

After the Agreement-Application has been signed, and login and password for the access to UBCH’s services have been obtained, a report can be requested in the following ways: