Biometrics

Computers can know who you are.

They look at your face. They look at your hands. They can even hear your voice. This helps keep things safe.

Some tools look at how you move. They might watch how you type. They might watch how you walk. This is a way to find you.

First, a machine must learn your traits. This is called enrollment. It saves your information in a list.

Later, the machine checks you again. It compares your body to the list. It looks for a match. This tells the machine it is you.

Some systems use more than one part of you. They might use your face and your hand. This can make the system even better.

Computers can identify people using biometrics. Biometrics are body measurements or ways of acting. These traits are unique to each person.

Some traits are about your body. These are called physiological traits. Examples include your fingerprints, face, or even your voice. Other traits are about how you act. These are called behavioral traits. This includes your typing rhythm or how you walk.

To start, a person must go through enrollment. This is when the system captures and saves your data. It creates a template. A template is a digital summary of your traits. Later, the system uses a matching phase. It compares your new data to the stored template.

There are two main ways to use these systems. Verification checks if you are who you say you are. It compares you to one specific person on file. Identification tries to find an unknown person. It compares you against a large group of people. Some systems use more than one trait. This is called a multimodal system. It can help make the system more secure.

Biometrics are special measurements of human features. They help computers identify people and control who can enter certain areas.

These identifiers are unique to every single person. This makes them more reliable than passwords or ID cards. However, collecting this data can raise questions about privacy. Scientists look at two main types of traits. Physiological traits focus on the shape of your body. This includes things like your fingerprints, face, or iris. Behavioral traits focus on how you act. This includes your typing rhythm or the way you walk.

Using these systems involves a specific way of working. First, a person goes through enrollment. During this step, the system captures and stores your information. It then creates a template. A template is a digital summary of your unique features. The system uses a sensor to get the data. Then, it cleans the data to remove background noise. Next, it extracts the most important features. These features are turned into a vector of numbers. Finally, the system performs a matching phase. It compares your new data to the stored template to see if they match.

There are two main modes for these systems. Verification mode is a one-to-one comparison. It checks if you are the specific person you claim to be. This might use a smart card or a username to help. Identification mode is a one-to-many comparison. It tries to find the identity of an unknown person. The system compares the sample against a whole database. If the match is close enough, the person is identified. Some systems even use negative recognition. This helps establish if a person is someone they deny being.

Researchers use several factors to pick the best trait. Jain et al. identified seven factors in 1999. One factor is universality, meaning everyone must have the trait. Uniqueness means the trait is different for everyone. Permanence means the trait stays the same over time. Measurability is how easy it is to collect the data. Performance looks at the speed and accuracy of the technology. Acceptability is how willing people are to use it. Finally, circumvention is how hard it is to fake the trait.

Sometimes, one trait is not enough for high security. This is why scientists use multimodal biometric systems. These systems use more than one sensor or trait at once. For example, a system might use both a fingerprint and a voice scan. This helps if one trait changes, like a cut finger. One complex idea is a multimodal cryptosystem. In a 2011 study, researchers used a face, fingerprint, and palm vein. The palm vein acts as a secret key. This makes the system very hard to trick or spoof.

Biometrics involve the measurement and calculation of human characteristics for identification. In computer science, this is known as biometric authentication. It provides a way to control access to systems or identify people within groups under surveillance. These identifiers are distinctive and measurable. They are often more reliable than traditional methods. Traditional methods include token-based systems like passports. They also include knowledge-based systems like passwords or PINs. However, collecting biometric data can raise significant privacy concerns.

Scientists categorize biometric identifiers into two main groups. The first group consists of physiological characteristics. These relate to the physical shape of the body. Examples include fingerprints, facial recognition, iris recognition, and DNA. Other examples are palm prints, hand geometry, and even the shape of ears. The second group includes behavioral characteristics. These are related to patterns of behavior. This category is sometimes called "behaviometrics." Examples include typing rhythm, mouse movement, gait, and signature.

To choose a biometric trait, researchers evaluate several specific factors. In 1999, Jain et al. identified seven key factors for assessment. Universality means every person must possess the trait. Uniqueness means the trait must distinguish individuals from one another. Permanence means the trait remains relatively invariant over time. Measurability, or collectability, refers to how easily the trait is acquired. Performance measures the accuracy, speed, and robustness of the technology. Acceptability is how willing a population is to use the technology. Finally, circumvention measures how easy it is to imitate a trait using an artifact.

A biometric system follows a strict technical process. It begins with enrollment, where information is captured and stored. First, a sensor acts as an interface to acquire data. This is often an image acquisition system. Second, the system performs pre-processing. This step removes background noise and artifacts. Third, the system extracts necessary features. These features are converted into a template. A template is a digital synthesis of relevant characteristics. It is often a vector of numbers or a specific image. Using a template reduces file size and protects identity. During the matching phase, a matcher compares the new template against stored ones. It may use algorithms like the Hamming distance to estimate the distance between them.

Biometric systems generally operate in two distinct modes. The first is verification mode, which is a one-to-one comparison. This process checks if a person is truly who they claim to be. It may use a username or a smart card to select a specific template. The second is identification mode, which is a one-to-many comparison. This mode attempts to establish the identity of an unknown individual. The system compares a sample against an entire database. If the match falls within a set threshold, the person is identified. This can include negative recognition. This helps establish if a person is someone they deny being. This is a function that passwords or keys cannot perform.

Many systems are unimodal, meaning they use only one biometric. However, unimodal systems have limitations. For instance, iris recognition can be affected by aging. Electronic fingerprint recognition can be hindered by cut or worn fingers. To solve this, scientists use multimodal biometric systems. These use multiple sensors or traits to overcome individual weaknesses. They can use different biometrics, such as a fingerprint and a voice scan. They can also use multiple images of the same marker. These systems can integrate information in several ways. They can use sequential, parallel, hierarchical, or serial integration modes.

Information in multimodal systems can be fused at different stages. Feature level fusion combines the actual data or extracted features. This is believed to be the most effective method. It provides richer information than other methods. Matching-score level fusion consolidates the scores from multiple classifiers. Decision level fusion combines the final results, such as through majority voting. Some advanced systems use a multimodal biometric cryptosystem. A 2011 study proposed a system using face, fingerprint, and palm vein. In this setup, the palm vein acts as a cryptographic key. This makes the system very difficult to forge or spoof. The system achieved 75% verification accuracy during testing. It required approximately 50 seconds for enrollment and 22 seconds for verification.