As an earlier post noted, advances in technology have spurred the implementation of various biometric authentication methodologies in the consumer market. But as people are discovering, not all methodologies are equally suited for all applications. Those who are implementing such applications have to consider risk level, cost, operating environment, and targeted population. They also have to evaluate a number of other factors to determine if a particular biometric is better suited than another for an intended application. These factors include but are not limited to:

  • Uniqueness. While the biometric doesn't always have to be unique to every individual on the planet, the probability that two people share a particular characteristic should be unlikely enough to prevent an unacceptable number of false acceptances (when one person is wrongly authenticated as another). For example, fingerprints are considered to be unique to every individual, but current smartphone fingerprint readers have such low-resolution scanners that the possibility of a false acceptance is one in 44,000. This rate is most likely sufficient for many applications, but a high-dollar transaction may require supplemental authentication.
  • Universality. The targeted characteristic must be present in the overall population, with only a few exceptions. Only a couple of biometric elements, such as DNA and facial recognition, can provide complete population coverage. Hand geometry and vein recognition, for example, won't work on people who are missing fingers or other body parts.
  • Permanence. The characteristic should not change over time. Even though people can alter almost any physical characteristic through medical procedures, the possibility of such alteration to the characteristic being considered for biometric authentication should be infrequent among the population—and the alteration procedure should be relatively expensive.
  • Collection ease. The more invasive the collection of the biometric sample, the more resistance people will have to it. People tend to view facial and voice recognition and fingerprinting as noninvasive but retinal scans as highly invasive—a light beam scans the back of the person's eye, which can be very uncomfortable.
  • Performance. The biometric element must support the creation of a template that is accurate and quickly obtained while also providing minimal database storage requirements. A system that takes a long time to authenticate someone during peak usage periods will encounter user dissatisfaction and possibly decreased productivity.
  • Accuracy. Individuals should not be able to fool the system. Fingerprint readers should verify that the right fingerprints belong to the right person, that a spoken phrase is live and not recorded, and so on.
  • User-embraced. Even when people have to use certain biometric authentication systems as a condition of their employment, the technology should be one that has a high level of acceptance, with minimal cultural, religious, collective bargaining, or regulatory implications.
  • Cost-effectiveness. As with all risk management practices, the cost of implementing and operating the system must be commensurate with the risk exposure for using a less secure authentication system.

As you consider the possibility of implementing a biometric authentication methodology for your customers, I hope you will find these evaluation elements helpful.

Photo of David LottBy David Lott, a payments risk expert in the Retail Payments Risk Forum at the Atlanta Fed