Types of Biometrics: Physical, Behavioural and How They Compare

You probably used a biometric today without thinking about it. Maybe a thumb on your phone or a glance at a banking app.

Those are just two of many types of biometrics. Some read your body. Others read your habits.

This guide walks through the main ones in plain language. You’ll see how each type works and where you meet it. You’ll also see how they compare when accuracy really matters.

Two families: what you are and how you act

A biometric is any trait a machine can measure to recognise you. The international vocabulary standard for biometrics puts it simply. It is recognition based on biological and behavioural characteristics.

That gives us two families. Physical biometrics (also called physiological) measure your body such as a fingerprint or an iris. Behavioural biometrics measure how you do things such as typing or walking.

The split matters in practice. Physical traits tend to stay stable for years. Behavioural traits shift with mood, injury or age yet can be checked quietly in the background.

So what are the 5 main types of biometrics?

It’s a common search. Yet there’s no official list of five. If you count what people use most, the answer is usually fingerprint, face, iris, voice and palm or hand.

That list leaves out a lot. Below we cover eleven types across both families so you can see the full picture.

Physical biometrics: reading the body

1. Fingerprint

The ridges on your fingertips form patterns that are unique to you. Even identical twins have different prints. A scanner maps the points where ridges end or split (known as minutiae).

Fingerprints are the most familiar type by far. You meet them on phones, laptops, office doors and border kiosks. Our guide to the types of fingerprints helps you find your own pattern.

2. Face

Face recognition measures the shape and layout of your features. Modern systems turn a face into a long string of numbers and compare that string with stored ones.

It’s the most convenient type because you only need to look at a camera. That same ease also makes it the most debated, since a camera can scan people who never agreed to it.

3. Iris

The iris is the coloured ring around your pupil. Its texture of furrows and freckles is highly detailed and stays stable for most of your life.

A near-infrared camera captures it and eye colour doesn’t matter. Iris scans are common in national ID schemes and some airports. We explain more in our guide to iris recognition.

4. Retina

The retina sits at the back of the eye. Its pattern of blood vessels is very hard to fake.

The catch is comfort. You need to hold your eye close to a scanner while light shines in. That’s why retina scans stay rare outside high-security sites.

5. Palm and finger vein

Vein scanners shine near-infrared light through your hand. Blood absorbs the light and the vein map shows up as a dark pattern.

Veins sit under the skin. That makes them hard to copy and less affected by cuts or dirt. Vein readers are popular at bank ATMs in Japan and in some hospitals.

6. Hand geometry

This older method measures the length and width of your fingers and hand. It’s quick and simple.

Hand shape isn’t unique enough to pick one person from millions. So it’s mostly used to confirm a claimed identity, like clocking in at work.

7. DNA

DNA is the most distinctive trait of all unless you have an identical twin. It’s also the slowest. A sample has to be processed rather than scanned.

That gap is closing. The FBI says its Rapid DNA booking process builds a profile from a mouth swab in about one to two hours. It still isn’t something you’d use to unlock a door.

Behavioural biometrics: reading your habits

8. Voice

Your voice sits on the border between the two families. The shape of your throat and mouth is physical. The way you speak is learned behaviour.

Voice checks are common in call centres and phone banking. The big debate now is AI voice cloning, which we explore in our piece on voice biometrics in banking.

9. Signature

A scanned image of a signature is easy to copy. Dynamic signature systems go further. They record the speed, pressure and rhythm of your pen as you sign.

You may have done this on a delivery tablet or a bank counter pad. It feels natural because people have signed things for centuries.

10. Keystroke dynamics

Everyone types with a rhythm. Keystroke systems measure how long you hold each key and the gaps between them.

It needs no extra hardware. Banks and online services often use it quietly to spot when someone else is typing on your account.

11. Gait

Gait recognition studies the way you walk. It can work from a distance and without your help. Few other biometrics can do that.

Accuracy drops when conditions change. A heavy coat, a bag or a different camera angle can throw it off. For now it’s mostly a research and surveillance tool.

All eleven at a glance

Type	Family	How it’s captured	Where you meet it
Fingerprint	Physical	Touch sensor or camera	Phones, laptops, borders
Face	Physical	Camera	Phones, airports, banking apps
Iris	Physical	Near-infrared camera	National ID, some airports
Retina	Physical	Close-up eye scanner	High-security sites
Palm and finger vein	Physical	Near-infrared light through the hand	ATMs, hospitals
Hand geometry	Physical	Hand reader	Workplace time clocks
DNA	Physical	Swab and lab or Rapid DNA unit	Police and forensics
Voice	Both	Microphone	Call centres, phone banking
Signature	Behavioural	Pen pad or tablet	Banks, deliveries
Keystroke	Behavioural	Your normal keyboard	Fraud checks in the background
Gait	Behavioural	Video	Research and surveillance

Which type is the most accurate?

The honest answer is that it depends on the test, the camera and the size of the database.

In 2004 a widely cited paper by Anil Jain and colleagues rated face recognition “low” for performance. Fingerprint, iris, retina and DNA were rated “high”. Two decades of deep learning have changed that picture.

What the independent tests show

The US National Institute of Standards and Technology (NIST) runs the best-known public tests. Each test uses its own data and settings, so the numbers below can’t be lined up as a league table.

• Face: in NIST’s September 2026 one-to-many results, the top algorithms missed the right person in about 0.05% to 0.1% of searches in some settings.
• Iris: NIST’s IREX IX study found the best two-eye matcher missed about 1 in 150 people in a database of 160,000.
• Fingerprint: in NIST’s PFT III test, the best algorithm in 2023 rejected the right finger 0.7% of the time while letting just 1 in 10,000 impostors through.

Behavioural types sit a step below. A 2025 review of keystroke research found error rates from under 0.1% to over 10%. It all depends on the dataset and the conditions.

Size changes everything

The chart shows why one accuracy number is never the full story. The typical system misses more people as a face database grows.

That’s why big national ID schemes rarely rely on one trait. Many combine face, fingerprints and iris. This approach is called multimodal biometrics.

Accuracy isn’t the only question

The “best” type also depends on what you’re willing to trade. Researchers often judge a trait on several points at once.

• Does everyone have it? Some people have worn fingerprints or eye conditions.
• Does it stay the same? Faces age and voices change. Irises and veins stay steadier.
• Will people accept it? Looking at a camera feels easy. Holding an eye to a retina scanner doesn’t.
• Can it be faked? Photos, masks and voice clones are real threats. Good systems add liveness checks to spot them.

These trade-offs explain why your phone uses face or fingerprint while a high-security lab might use iris or veins. We look at them in more depth in our guide to biometrics pros and cons.

Where the debate gets heated

Biometrics bring real convenience. They also raise questions that experts still argue about.

You can’t reset your body. A leaked password can be changed. A leaked fingerprint template can’t.

Supporters reply that modern systems store a mathematical template rather than an image. On phones that template never leaves the device.

Some types work better for some people. NIST tested 189 face algorithms in 2019 and found that most made more errors for certain demographic groups. It also noted that the fairest algorithms ranked among the most accurate.

Critics say any gap is too much in policing. Vendors point to steady progress since then.

Consent depends on the type. You choose to press a fingerprint reader. A face or gait system can scan you from across a room.

Some see that as a key safety tool. Others see it as surveillance by default.

Where you stand on these depends on the setting and on who holds the data. A face unlock on your own phone is a very different thing from a camera in a public square. If you want to see how these traits are used to prove who you are, our guide to biometric authentication goes step by step.

Short answers to common questions

What are 3 examples of biometrics?

Fingerprint, face and iris are the three you’re most likely to meet. All three are physical biometrics.

What is the difference between physical and behavioural biometrics?

Physical biometrics measure a part of your body such as a fingerprint. Behavioural biometrics measure how you act. Think of how you type or walk.

What types of biometric data are there?

There are images such as a face photo or a fingerprint scan. There are also recordings such as voice and video plus measurements such as typing rhythm. Most systems keep a numeric template rather than the raw image.

Which biometric is the hardest to fake?

Veins and the retina sit out of sight and are the hardest to copy. Iris is close behind. Face and voice are easier to imitate without liveness checks.

How many types of biometrics are there?

There’s no fixed number. This guide covers eleven common types. Researchers keep testing new ones such as ear shape and heartbeat patterns.

Your body, your call

Every type of biometric is a trade between convenience, accuracy and privacy. Fingerprints and faces win on convenience.

Iris and veins win on stability. Behavioural traits win on working quietly in the background.

So which would you trust most with your bank account? Would you let a camera recognise you in a shop if it saved you time?

And should any of these traits be used without your knowledge? There’s no single right answer. There’s only the one that fits you.

Sources

• ISO/IEC 2382-37:2022, Information technology vocabulary, Part 37: Biometrics
• Jain, Ross and Prabhakar (2004), An Introduction to Biometric Recognition, IEEE Transactions on Circuits and Systems for Video Technology
• NIST: Face Recognition Technology Evaluation (FRTE) 1:N Identification
• Biometric Update: New NIST FRTE 1:N results illustrate facial recognition accuracy’s multidimensionality (September 2026)
• NIST IR 8207 (2018): IREX IX Part One, Performance of Iris Recognition Algorithms
• Biometric Update: Neurotechnology fingerprint biometric accuracy reaches new high in PFT III (April 2023)
• NIST (2019): NIST study evaluates effects of race, age, sex on face recognition software
• Shadman et al.: Keystroke Dynamics: Concepts, Techniques, and Applications
• FBI: Rapid DNA