The main questions for implementing a biometric security system are: Is it practical (i.e., can it deal with large numbers of personnel in a timely manner and achieve security goals)? Is it efficient (i.e., Can it verify someone quickly without a large amount of time in training, and is the system flexible enough to meet your needs?)? And is it cost-efficient (i.e., Can it be implemented quickly and inexpensively and what is the ROI over a period of 2-5 years?)?
Q: Where has facial biometrics implementation already seen its greatest successes?
A: Mainly in physical access control implementations and law enforcement. The problem of "token passing" between individuals just points out the need for a higher level of authentication and facial biometrics is regarded by most users to be the least intrusive of all of the biometrics. Several of our installations have been in operation for over a year with a zero percent false acceptance rate. In law enforcement, not everyone has fingerprints on file, and sometimes all they have to go on is a face. The ability to search for a criminal by their face alone will ensure that criminals will not be released by using false identities and/or making bail before law enforcement officers find out who they really are.
Q: Some of us aren't the technical engineers we wish we were - can you briefly summarize the different types of facial recognition technology and briefly (and in laymen's terms) explain how they work?
A: There are two main camps in the facial biometric arena - 2D and 3D.
2D systems generally utilize Eigenfaces to recognize faces. This technology was invented in 1987 by a couple of engineers at MIT and is considered to be the first facial working facial recognition technology. This method takes a large number of points (example: from the tip of your nose to the left edge of your mouth, from the tip of the nose to the outside of the right eye, etc.) and compares commonalities and differences between groups of individual facial images. The Eigenfaces method has difficulty when presented with different light levels and pose positions. In addition, the face must be presented to the system as a frontal view in order for the system to work. The 2D systems can usually be fooled by a digital photograph or digital video.
The 3D systems are generally based on a neural network. These systems are much more advanced in being able to process information at a much higher speed, with greater accuracy than 2D systems. In addition, these systems have the ability to learn, which makes the issues of aging, poses, glasses and facial hair a relatively insignificant objection. Since the neural network learns from experience it does a better job of differentiating in varying light conditions and greatly improves accuracy over 2D technology. 3D systems can tell identical twins apart and cannot be fooled by digital photographs or digital video.
Q: Biometrics keeps gaining momentum, but token-based access control systems are here to stay. How do these systems (facial biometrics and card-type access systems and IT network access control) work together in an ideal world?
A: In an ideal world, an organization already has an access control solution using proximity cards. The addition of facial biometrics merely adds a second level of authentication to ensure that the token holder is the registered owner of the token and not someone else trying to use it. Typical card access systems will keep a record of who comes and goes. However, it only lists the name of the person who is registered to the token. With the addition of facial biometrics, you not only have a record of who comes and goes, but a real time image of the person requesting access as well. In addition, card access systems will let anyone in as long as the card is valid, and some systems even have a digital photo of the registered owner. However, this requires a visual inspection of the photo against the person utilizing the card. Facial biometrics not only removes this manual inspection process, but it doesn't sleep, doesn't take breaks and can't be distracted.
Q: What about the issue of privacy in using facial biometrics?