Biometrics

Biometric systems operate by capturing a trait, extracting distinctive features, and comparing them against a stored template. The two main categories are physiological traits (fingerprint, face, iris, palm, DNA) and behavioral traits (voice, gait, keystroke dynamics). Each modality has trade-offs in accuracy, user acceptance, and spoof resistance. For example, iris recognition is highly accurate but requires cooperative subjects and specialized cameras, while face recognition is non-intrusive but can be affected by lighting and aging.

Performance is measured by false acceptance rate (FAR) and false rejection rate (FRR), which are inversely related. A system's equal error rate (EER) is the point where FAR and FRR are equal, often used to compare systems. Multimodal biometrics, which combine two or more traits, can reduce error rates and improve robustness against spoofing.

The use of body measurements for identification dates back to Alphonse Bertillon's anthropometric system in the 1880s, which measured bones and body dimensions. Fingerprinting was adopted by law enforcement in the early 20th century, replacing Bertillonage due to its higher reliability. Automated biometric systems emerged in the 1960s with work on face recognition by Woodrow Bledsoe and Helen Chan Wolf, and the first commercial fingerprint scanners appeared in the 1970s.

The 1990s saw the rise of iris recognition, pioneered by John Daugman, whose algorithms remain the basis of most iris systems today. The 2000s brought biometric passports and large-scale national ID programs, such as India's Aadhaar, which has enrolled over 1.3 billion people. Recent advances in deep learning have dramatically improved face recognition accuracy, enabling widespread use in consumer devices and surveillance.

Biometrics are used in law enforcement (fingerprint databases, facial recognition for suspect identification), border control (e-passports, automated gates), and consumer electronics (fingerprint sensors, Face ID). Financial institutions use voice and fingerprint biometrics for fraud prevention, and healthcare uses palm vein recognition for patient identification. Behavioral biometrics, such as keystroke dynamics, are increasingly used for continuous authentication in cybersecurity.

Large-scale deployments include the FBI's Integrated Automated Fingerprint Identification System (IAFIS) and the European Union's Entry/Exit System (EES), which will record biometric data of non-EU travelers. Biometric voter registration has been used in several African countries to reduce electoral fraud. However, failures can have serious consequences: false matches can lead to wrongful arrests, as seen in a 2019 case in the United States where a man was falsely accused based on a facial recognition match.

Biometrics extend beyond common modalities: ear shape, odor, and even brainwave patterns have been studied as identifiers. The human ear has a unique structure that remains stable with age, making it useful for surveillance in some contexts. Gait recognition can identify individuals from a distance, but it is highly sensitive to clothing and surface. Keystroke dynamics, which measure typing rhythm, are used for continuous authentication but can be affected by injuries or changes in keyboard.

Biometric data is irrevocable: unlike passwords, a compromised fingerprint cannot be changed. This has led to the development of cancelable biometrics, which transform raw data into revocable templates. Another niche area is biometric liveness detection, which aims to distinguish real traits from spoofs, such as silicone fingers or printed photos. The field also intersects with forensics: latent fingerprints can be enhanced with chemical methods, and DNA phenotyping can predict physical appearance from genetic material.

False acceptance rate (FAR)

The probability that a system incorrectly matches an input to a non-matching template.

False rejection rate (FRR)

The probability that a system fails to match an input to a matching template.

Equal error rate (EER)

The point at which FAR and FRR are equal; lower EER indicates better accuracy.

Multimodal biometrics

The use of two or more biometric traits to improve accuracy and security.

Cancelable biometrics

A technique that transforms biometric data into a revocable template to protect privacy.

Liveness detection

Methods to ensure that a biometric sample is from a live person, not a spoof.