Facial recogntion
By 2008 it had already been recognized that facial recognition was a flawed biometric.

Data acquisit
ion, usually in the form of a photograph, while appearing to be a simple process, is not. Both the FRR and the FAR are unacceptably high.

The gap between the myth and reality of the technology is well encapsulated in the article about the Boston bombings. Facial recognition software was unable to recognize the bombers, although the CCTV footage proves that they were there.

The technology has been dropped by the DHS because it has been proven to impact negatively on the ability of immigration officers to observe other characteristics that suggest that a person might be undesirable.

Facial features change over time. This exaggerates the error rates, making it difficult to determine with certainty that Chris Beck (left) and Kristin Beck (right) are the same person.

  • No contact required
  • Commonly available sensors (cameras)
  • Large amounts of existing data to allow background and/or watchlist checks
  • Easy for humans to verify results

  • Face can be obstructed by hair, glasses, hats, scarves, etc.
  • Sensitive to changes in lighting, expression, and pose
  • Faces change over time
  • Propensity for users to provide poor-quality video images yet to expect accurate results

Facial recognition is highly effective for narrowing down possible matches in large crowds. It is less effective that other modalities at definitively identifying a person.

More at:
Assessing Face Acquisition
Biometric data interchange formats
Clocking people’s clocks
Doctor 'used silicone fingers' to sign in for colleagues
European Union – Biometric Matching System
Face Recognition
Facial-Recognition Technology Proves Its Mettle
Live Face Detection
Overview of the Multiple Biometrics Grand Challenge
Performance of Fingerprint Match-on-Card Algorithms Phase II / III Report
Photograph guidelines