Skip to content

Introduction

Face biometrics is a state-of-the-art technology allowing to validate a person identity by means of his face. Veridas solution, das-Face, captures the unique features and characteristics of a face, generating a face biometric descriptor characterizing uniquely the person.

The facial biometric descriptor is a mathematical representation obtained from the specific set of features found in a person face. The conversion from face image into a biometric descriptor is technically irreversible, so it is not possible to recover a person’s face from the resulting descriptor. This descriptor has many advantages, among others, requires much less storage than the original image.

The solution developed by Veridas and implemented in das-Face product is a service-oriented architecture that can be consumed via APIs.

Algorithm performance is shown on the document Face Biometry Performance Report

das-Face computes the similarity between faces recorded in images among other operations. The main operations implemented in this product are:

  • Verification (1:1 matching): This operation compares two face images to find out that they belong to the same individual.
  • Liveness: This operation computes how likely a given capture contains a genuine face, or if contrary, the same is the result of a spoof attempt (for intance presenting the picture of a face instead the face itself).
  • Credential generation: This operations returns a biometric descriptor computed for the face represented in a given image. This biometric descriptor is encoded and ciphered into a proprietary biometric credential format.

Veridas active liveness detection implemented in Selfie-Alive Pro was tested by iBeta to the ISO 30107-3 Biometric Presentation Attack Detection Standard and was found to be in compliance with Level 1 and Level 2.

Alt

das-Face is offered as a REST API. The most common process to obtain the similarity between two faces is:

  1. Two images are sent to the API.
  2. Both images are pre-processed, detecting the faces location in both images, and normalizing the detected faces in the way required for the next process.
  3. The normalized face image is converted into an irreversible mathematical descriptor (facial biometric descriptor).
  4. Both vectors are compared and a matching score between 0 and 1 is provided. The matching score represents a metric of the similarity between both face images. The higher the score, the greater the certainty to be the same person.
  5. You can use this matching score to validate the identity of a customer. It is recommended to define a threshold within required confidence level using the FPR (False Positive Rate) and FNR (False Negative Rate) expected ratios. The calibration curves are shown at the document Face Biometry Performance Report.

Veridas does not store any personal data during the operation of this product. All the user information is immediately deleted.