Biometrics are measurements of human characteristics. They are distinctive, cost effective to acquire and compare, and used to help enroll, identify, verify and authenticate people. There are other important aspects of a good biometric, such as stability. Biometric stability is the desirable attribute that persists over a long time frame. If fingerprint friction ridges changed from one day to the next, we would not be able to use them so readily. For many use cases, such as forensic analysis, border control and physical access control, we need a biometric that is stable for decades.
So – Are Our Biometrics Stable?
The short answer is yes. If they were not stable, then the dominant biometrics that we see in use today would not be as well adopted. Fingerprints stay the same for decades, as do iris muscular structure, vascular structures and bone structures that account for the usability of vein recognition, speaker recognition, hand geometry or facial recognition. Even the physical characteristics that make a voice distinctive are stable for long periods of time. Even so, biometric stability is not perfect as we age. As we mature from small children into adults and as we lose our youthful regenerative capacities, our physiology changes and therefore so do our biometrics. Additionally, as we live, we continue to aggregate exposure to an environment that we cannot fully recover in our healing processes. Scars can develop on our skin that impact fingerprint and facial recognition, prolonged exposure to radiation can induce cataracts, etc.
There is a significant amount of past and present research into these subjects helping us understand the limits of biometric stability and informs us about realistic expectations about the use of biometrics for application use cases. In the academic literature, these are often called longitudinal studies, as they often acquire biometrics from the same group of people along a significant period of time spanning years or decades.
A Look at the Findings
Fingerprints are the most adopted biometric technology. There have been significant studies about the stability of fingerprints. The friction ridge pattern used for almost all fingerprint recognition systems is well set before birth (roughly at the age of 6 months in the womb), but as children’s fingerprints grow, the scale of the friction ridges changes, such that ridges that were once close together grow apart. As the skin is an elastic layer above a complex and unique 3D shape of a finger, not only does the scale change, but complex nonlinear transformations come into process as the child matures. What this means is that if you enroll a fingerprint of a child, you can only use it for a short period of time. Performance begins to rapidly fall off after perhaps 4 years (based on a recent study led by Dr. Gunter Schumacher funded by the European Commission Joint Research Centre). After a child is about 12 years old, however, the fingerprints are stable enough to use modern matchers for long periods of time. Still, due to issues of exposure, it is good practice to re-enroll or update reference biometric material from time-to-time.
A permanent scar will have a negative impact on recognition rates, but the scar can often be a distinguishing characteristic upon a re-enrollment. Passports for instance are only good for 10 years, for a variety of reasons, and renewals require a collection of fresh fingerprints. As we get old, our fingerprints show the wear of many years. Our pores aren’t keeping our ridges lubricated as well which results in low quality fingerprints. While the fingerprint ridges of old people are still substantially similar to the ridges collected when they were young, the poor quality can make fingerprint matching a challenge. Exactly when this happens varies by individual. In any case, it is wise to have an enrollment quality threshold so that an identity management system can rely on poor quality biometric characteristics.
Face biometrics are another significant biometric. While we might know of them from their dominant role in passport verification at borders, they are increasingly used for social media tagging, logical access control and identification from video surveillance systems. Similar to fingerprints, faces rapidly change as we mature from children into adults. Faces change the fastest from birth until 3 years old, with another rapid change from 13 to adulthood. Furthermore, we see significant changes in faces as we age from about 40 years on. The consensus advice is that faces should be re-enrolled every 5 years. However, recently the capabilities of facial recognition are improving at a very rapid rate. There have been advances in predicting the effects of aging which allows for significant improvement in accuracy.
A final biometric we might consider is our voice – utilized in speaker recognition. This is a very difficult topic to study, as several other factors such as health, emotional state, background noise, microphone quality and channel characteristics all likely have a larger impact than aging. Nevertheless, as a voice print relies on the physiological characteristics of a person’s vocal tract, changes in that vocal tract as we age will correlate to degradation in recognition accuracy. Like with facial recognition, speaker recognition systems may be able to predict changes and improve performance. While we know of definitive guidance to the longevity of voice prints, we recommend that voice prints be re-enrolled every 5 years.
Biometric Stability and System Performance
The use of biometrics is a valuable part of an identity management solution, as it combines good security with strong convenience aspects. Almost everyone has unique fingerprints, a face and a voice, and they are cost effective to acquire and compare. These are also relatively stable, such that there is a strong return on the investment of a system and the enrollment of the users over the expected lifetime of the managed identities. It is important to understand the limits of biometric stability. Continued research will not only give us better policy guidance, but will inform improved algorithms and lead to more robust performance. While we recommend that biometric systems establish an aging aware policy, we are confident that aging aspects are not a primary factor in the performance of biometric systems.
Greg Cannon is Vice President & Chief Technology Officer of Crossmatch responsible for the Company’s standards involvement, intellectual property, software architecture, biometric algorithm development, and continued innovation excellence.