New Face Morphing Dataset (for vulnerability research)

 

Biometix is pleased to announce our freely available Facial Morphing Dataset. Morphing produces facial images with the features of two (or more) individuals. This a vulnerability in the context of any ID issuance (but particularly Passport and border operations) as it can allow multiple people to have an ID that would pass both visual and biometric checks.

Previous studies have found that morphed samples return high matching scores against biometric algorithms commonly used in automated border control. To help assess this issue and produce standardized test results Biometix has produced a test set of standard morphed facial images based from the well known NIST FERET dataset (https://www.nist.gov/itl/iad/image-group/color-feret-database). These images may be used for vulnerability testing of biometric systems.

The Biometix morph set can can answer two key questions:

  • Are the biometric scores from morph images high enough to be accepted by any given face recognition engine?
  • Are morphs able to be detected via vulnerability detection (or presentation attack detection – PAD) measures?

By undertaking this study biometric system owners can understand the level to which their biometric system is vulnerable to these morph attacks. We also hope it will spur on going development in detection techniques.

If you would like access to the morph set or to hear more about Biometix’s projects in biometric vulnerabilities, please contact us.