Biometric identification systems

DEFINITION: Technologies that use automated measurements and database comparisons of physiological and behavioral characteristics to identify target individuals.

SIGNIFICANCE: Biometric identification systems are becoming increasingly important given heightened concerns with security in many contexts. Compared with many other means of authorization and authentication, including password recognition, biometric technologies represent a significant advance in terms of ease of use, reliability, and validity.

The constantly evolving science of biometrics has produced a wide variety of systems capable of comparing hand, facial, eye, signature, vocal, brain, and genetic measures of given individuals against profiles of such measures stored in large databases. The applications of this technology for law-enforcement purposes are extensive. Biometric systems have been used to identify offenders who are using aliases, to fight illegal immigration, and to identify inmates as they are moved through various phases of the correctional system. Biometric data can be used to verify identity claims or to screen for persons who have been identified as potential security risks.

Accuracy

Biometric identification systems represent a huge improvement over the traditional “token” (credit card or document) and password systems. Credit cards can be lost or stolen and then used as false identification. Similarly, passwords can be “cracked,” forgotten, or stolen. Biometric characteristics, on the other hand, are much more stable and permanent. Their inherent complexity renders them difficult or impossible to replicate, and the person being identified usually needs to be physically present at the time of the verification attempt. In addition, biometric systems can couple identifying information with other important background data, such as health or employment records (a fact that has led some to criticize the use of these systems as infringing on civil liberties).

The components of the typical biometric system are relatively straightforward; they consist of a sensor and a computer. The sensor is the device that gathers the biometric data from the individual being evaluated. The computer then processes the data collected; in some cases, the computer may refine the data by removing irrelevant information and background “noise” that may interfere with the interpretation of the results. The computer captures the biometric features being measured and creates a template, which it then compares to a database of biometric information on known individuals, looking for an identification match, or “hit.” The consequences of a successful identification are as varied as the systems themselves. At the point of identification, an individual might be allowed into a restricted area, picked up for further questioning in a specific investigation, or observed further for any suspicious behavior.

The accuracy of a biometric system is typically assessed using one or more of the following measures: the failure-to-acquire rate (a measure of the percentage of unsuccessful attempts by the system to obtain specific biometric information from subjects), the false-accept rate (also known as the false-positive rate, a measure of the percentage of incorrect matches of subjects’ biometric profiles to profiles already included in the database), and the false-reject rate (also known as the false-negative rate, the percentage of failures to match subjects’ biometric profiles with identical profiles already included in the database). Minimization of all these kinds of error rates reduces the numbers of suspects who are needlessly detained, restricted from air travel, or otherwise affected by law-enforcement “false alarms” while maximizing the appropriate identification of true security threats.

Applications

Law-enforcement agencies employ biometric technologies in many ways, including for facial recognition, fingerprint identification, iris recognition, and voice recognition. Facial recognition systems use specific aspects of facial features from scanned photographs to make identifications. The features analyzed may include the physical distance between specific features, skin color, thermal patterns of blood flow, and facial lines. One application of facial recognition technology is the establishment by departments of archives containing many thousands of offender photographs. These are matched with suspects’ pictures or used to produce photo lineups that can be shown to crime victims or witnesses.

Numerous evaluations of facial recognition technology have produced mixed results. One Australian system, for example, tested in the Sydney airport, was found to have a false-reject rate of 2 percent. This rate was confirmed by tests sponsored by the US government. Although this error rate seems low, major world airports typically service several million passengers annually, which means that the systems could potentially falsely reject many thousands of people. One meta-analysis of facial recognition systems produced accuracy rates ranging from 51 percent to 94 percent. Factors affecting the rates included lighting, the quality of the photographs taken, movements of the subjects, the angles of the poses in the photographs, and the presence of eyeglasses on subjects. In general, male subjects and older persons were more easily recognized than were female and younger subjects. An inverse relationship was also found between accuracy and the size of the database against which the subjects’ facial features were compared.

Fingerprint identification is the oldest form of biometric identification, having been in use since the late nineteenth century. The Federal Bureau of Investigation (FBI) established a central database of fingerprints in 1924 against which law-enforcement agencies can seek to match the prints of crime suspects and victims. With modern electronic and laser technology, fingerprint images are often taken and transmitted “live” to the database. Efforts to automate the analysis and identification of fingerprints began in the 1960s.

Fingerprint identification systems use electronic fingerprint readers to locate where the ridges of fingerprints start, end, or split up. These areas, known as minutiae points, form the basis for the identification. Each fingerprint typically contains thirty to forty minutiae points, and no two people’s prints will match on more than eight such points.

In terms of accuracy, the false-accept rates of fingerprint identification have generally been very close to zero, and false-reject rates have been 3 percent or less, according to a 2015 study by the Miami-Dade Police Department and the Office of Justice Programs. The accuracy of the analysis of fingerprints taken from crime scenes, however, is often reduced because of the poor quality of the prints themselves. A 2020 study by Forensics Resources found that, as the size of fingerprint databases continues to grow, the number of close non-matches (CNMs) could eventually be as high as 28 percent.

In addition, although it is often assumed that fingerprints are stable over a lifetime, research has shown that they in fact can change in response to physiological growth, activity, or intentional alteration; it has also been shown that many fingerprint matching systems can be “spoofed.” Despite some limitations, fingerprinting is less controversial and more highly developed than any other type of biometric identification system. This is reflected in court acceptance of fingerprinting evidence. Moreover, fingerprint identification is no longer the sole purview of law enforcement and military forces; by 2014, commercial fingerprint-scanning modules were available for the protection of business and consumer computers as well as smartphones.

In iris recognition identification systems, an image of the iris of the eye (the colored ring surrounding the pupil) of the person to be identified is recorded by a digital camera and then converted into a template, which is checked for matches against an existing database. False-positive rates for such systems have averaged 0.1 percent, and false-negative rates have averaged 1.5 percent. An advantage of using this biometric technique is that, unlike fingerprints, the structure of the iris is permanent by the age of one and is unique for each person (this includes comparisons between identical twins and even between the left and right eyes of the same person). Unlike with fingerprint identification, however, iris evidence is not left at crime scenes. In addition, failure rates as high as 15 percent have been found when iris-scanning technology is used in brightly lit settings. This technology has many potential applications, including security screening at airports and borders, passport and immigration control, and identification for banking and issuance of drivers’ licenses.

Voice recognition systems use physical and behavioral aspects of the voice to identify individuals; the voice features measured are based on the physiology of the windpipe, nasal cavity, and vocal cords. A digital “voice signature” is recorded, and a computer measures the features and compares them against known samples for identification and verification. One drawback to the use of voice biometrics is that voice patterns can vary with age, and they can also be affected by medical problems (including even a cold) and the emotional state of the examinee. Background noise can also be a problem with the use of this identification technology.

DNA identification relies on an individual's unique “variable number tandem repeats (VNTRs)” within that person's DNA. VNTR sequences are similar for relatives and the same for identical twins but unique among strangers. Thus, analysis of VNTRs can help scientists identify individuals and relatives.

Another biometric identification technology that has been investigated is hand geometry scanning, which involves more than ninety measurements of different parts of the hand. To detect forgery, dynamic signature identification has been developed; in this system, the specific dimensions of the pen strokes a person makes while writing his or her signature (including pressure, speed, and direction) are recorded and stored for later matching. This technology is prone to high false-negative rates, however, because even though signatures are ubiquitous in daily transactions, only specific parts of a person’s signature remain constant across every signing. Gait analysis, which focuses on people’s unique walking patterns, is another type of biometric technique. Limitations to gait analysis include the fact that making gait measurements may be invasive; also, gait can be affected by injury or by a change in shoes. Researchers in 2023 were assessing the value of photogrammetry, which can be used by medical examiners to recontruct the geometry and body surface of deceased persons, creating a 3D model. They found this procedure to be more accurate than hand geometry scanning.

Bibliography

Crompton, Malcolm. “Biometrics and Privacy: The End of the World as We Know It or the White Knight of Privacy?” In Biometrics: Security and Authentication. Sydney: Biometrics Institute, 2003.

Das, Ravindra. Biometric Technology: Authentication, Biocryptography, and Cloud-Based Architecture. CRC Press, 2015.

Federal Bureau of Investigation. "Next Generation Identification (NGI)." Fingerprints & Other Biometrics, 2016.

"Fingerprint Examiners Found to Have Very Low Error Rates." PR Newswire US, February 2, 2015.

Jain, Anil K., Arun Russ, and Sharath Pankanti. “Biometrics: A Tool for Information Security.” IEEE Transactions on Information Forensics and Security 1, no. 2 (2006): 125–143.

Krishnan, K. N., with D. R. Berwick. Developing a Police Perspective and Exploring the Use of Biometrics and Other Emerging Technologies as an Investigative Tool in Identity Crimes. Payneham, S. Aust.: Australasian Centre for Policing Research, 2004.

Mansfield, A. J., and J. L. Wayman. Best Practices in Testing and Reporting Performance of Biometric Devices: Version 2.01. Teddington, Middlesex, England: National Physical Laboratory, 2002.

Mills, Kelly. “University Opts for Biometric Security.” Computerworld, January 25, 2002, 3–4.

Olson, Sarah. "Fingerprint Error Rate on Close Non-Matches." Forensics Resources, Office of Indigent Defense Services, 7 Oct. 2020, forensicresources.org/2020/fingerprint-error-rate-on-close-non-matches/. Accessed 14 Aug. 2024.

Ujvari, Zsolt, Mate Metzger, and Gergely Gardonyi. "A Consistent Methodology for Forensic Photogrammetry Scanning of Human Remains Using a Single Handheld DSLR Camera." Forensic Sciences Research, vol. 8, no. 4, Dec. 2023, pp. 295-307, /doi.org/10.1093%2Ffsr%2Fowad036. Accessed 14 Aug. 2024.

Vacca, John R. Biometric Technologies and Verification Systems. Burlington, Mass.: Elsevier, 2007.