University of Padova

Department of Information Engineering - DEI

Biometrics
Biometric identifiers

Stefano Bonaldo

Introduction 1
 Biometric trait

Given the multitude of characteristics coupled to a human being, we need to classify the
biometric traits in two main groups:
1. Physiological traits are related to uniquely measurable physical traits of a person
2. Behavioral traits are acquired over time and are related to measurable patterns in
human activities

Introduction 2
 Biometric identifier

A human anatomical or behavioral trait becomes a biometric identifier when it recognizes a

person by satisfying the following requirements:
▪ Universality
Each individual possess that biometric trait
▪ Uniqueness
The biometric trait is different in each person
▪ Permanence
The trait is invariant with respect to the matching criterion over time
▪ Collectability
The measure of the trait is possible to collect in an easy and comfortable way
▪ Measurability
The trait can be measured quantitatively

Introduction 3
 Important factors

In practical biometric systems, we need to select a suitable measurable identifier

depending on the application and to consider the following important factors:
▪ System performance
The performance includes recognition accuracy, speed (throughput), resource
requirements, and robustness to operational and environmental factors.
▪ Social acceptability
The acceptability evaluates if users are willing to accept the biometric identifier and how
much comfortable is its use in their daily lives
▪ Circumvention level
This factor evaluates how easy the biometric system can be circumvented by fraudulent
methods
A biometric system should have acceptable recognition accuracy and speed with reasonable
resource requirements, harmless to the users, accepted by the intended population, and
sufficiently robust to various fraudulent methods.
Introduction 4
 Selection of the identifier - I

Different types of biometric identifiers are in use and each of them has its own strengths
and weaknesses, depending on the application. No single identifier is expected to effectively
meet the requirements of all the applications. The match between a biometric trait and an
application is determined depending upon the characteristics of the application and the
properties of the trait. When selecting a biometric identifier for a particular application, we
need to evaluate the following points:
▪ Does the application need a verification or identification system? If an application
requires an identification of a subject from a large database, it needs a very distinctive
biometric trait, e.g., fingerprints or iris.
▪ Definition of operational characteristics of the application: is the application attended
(semi-automatic) or unattended (fully automatic)? Are the users habituated (or willing to
become habituated) to the given biometric? Are subjects cooperative or non-
cooperative?

Introduction 5
 Selection of the identifier - II

▪ What is the template storage requirement of the application? For example, an application
that performs the recognition on a smart card may require a small template size.
▪ How stringent are the performance requirements? For example, an application that
demands very high accuracy needs a more distinctive biometric.
▪ What types of biometric identifiers are acceptable to the target users? Biometric traits
have different degrees of acceptability in different demographic regions depending on the
cultural, ethical, religious, and hygienic standards. The acceptability of a biometric in an
application is often a compromise between the sensitivity of the targeted population to
various perceptions and the value or convenience offered by biometrics-based recognition

Introduction 6
 Biometric identifier: fingerprint - I

Fingerprint recognition
Fingerprint recognition is the most common biometric identifier in
use. A fingerprint is the pattern of ridges and valleys on the surface
of a fingertip whose formation is determined during the first seven
months of fetal development. It has been empirically determined
that the fingerprints of identical twins are different and so are the
prints on each finger of the same person.

Introduction 7
 Biometric identifier: fingerprint - II

1. Optical sensors 2. Solid-state sensors

▪ FTIR ▪ Capacitive ▪ Ultrasonic
▪ FTIR variants ▪ MEMS ▪ Electro-optical
▪ Multispectral ▪ Pressure
▪ Thermal

Introduction 8
 Biometric identifier: fingerprint - III

Compared to other recognition modalities, the fingerprint is very

attractive biometric identifier because:
▪ Each fingerprint has a very high level of uniqueness
▪ The fingerprint normally stays permanent during the whole
lifespan of a person
▪ Existing technologies allows for a cost-efficient sensing and
measurement of the fingerprint
▪ The sensed parameters of the fingerprint can easily be quantified
allowing efficient algorithms to identify the user
▪ Many standards relating to fingerprint recognition have been
already defined and have created the basis for an economy of
scale, i.e. have lowered the cost of fingerprint hardware and
software

Introduction 9
 Biometric identifier: fingerprint - III

Fingerprint recognition gets high performance scores in terms of

ease of use, processing speed and general security. Hence
fingerprint recognition comes out as an excellent biometric
identifier in most situations, and especially for high volume
NORMAL applications where cost and standardization are of particular
importance. Other biometric identifiers can be used as
complements, when security requirements demand multimodal
authentication of a person identity.
DRY Finger condition, e.g. dirty or dry fingers, can affect the sensing,
causing the recognition process to fail for some sensor types.

WET
Introduction 10
 Biometric identifier: face

Face recognition
Face recognition is a non-intrusive method, and facial
attributes are the most common biometric features used by
humans to recognize each others. Applications of facial
recognition range from a static, controlled “mug-shot”
authentication to a dynamic, uncontrolled face identification in
a cluttered background. Approaches to face recognition are
based on:
▪ Location and shape of facial attributes, as the eyes,
eyebrows, nose, lips, and chin, and their spatial locations.
▪ The overall (global) analysis of the face image that
represents a face as a weighted combination of a number
of canonical faces.

Introduction 11
 Biometric identifier: face recognition

The authentication performance of face recognition systems depends

on their application, as they can impose a number of restrictions on
how the facial images are obtained. These systems have difficulty in
matching face images captured from two different views, under
different illumination conditions, and at different times. It is
questionable whether the face itself, without any contextual
information, is a sufficient basis for recognizing a person from a large
number of identities with an extremely high level of
confidence. In order for a facial recognition system to
work well in practice, it should automatically:
▪ Locate the face position in the acquired image
▪ Recognize the face from a general viewpoint (i.e.,
from any pose) under different ambient conditions

Introduction 12
 Biometric identifier: face recognition

Pros
▪ Low cost - A common CCD/CMOS image camera is enough
▪ Range - Depending on what you want to achieve, facial recognition has an advantage in
that it can be used at rather long distances and without a person knowing that he or she
is being identified.
Cons
▪ Influence from the environment condition- Requires good ambient lighting
▪ Low stability over time – Face changes with age as well as disease and weight loss/gain.
▪ Medium security – It can be easy to spoof a facial recognition system by altering the face
with prosthetics, often a photo will do the job.
▪ High failure rate – glasses, hats, haircuts and a myriad of other factors can make facial
recognition system unable to capture the required nodal points needed for
authentication (masks during COVID emergency).
Introduction 13
 Biometric identifier: example of applications

Protect law enforcement - Mobile face recognition helps police officers to instantly
identify individuals in the field from a safe distance. This allow giving them contextual
data of the person in front of them. As an example, if a police officer pulls over a
reported person at a routine traffic stop, the officer would instantly know that the
person may be dangerous or armed and could call for reinforcement and be aware of it.
Find missing people, e.g. children and victims of human trafficking - As long as missing
individuals are added to a database, law enforcement can become alerted as soon as
they are recognized by face recognition in airports, retail stores or other public spaces.
In India, 3000 missing children were discovered in just four days using face recognition.

Track school attendance - In addition to school security, face recognition is currently

used to track the presence of students in classes, e.g. in China and Australia.
Traditionally, attendance sheets are time consuming for the teacher and may be
unreliable in crowded classes, as a student can sign for another student.

Facilitate secure transactions - In China, there is a financial services company called

Ant Financial that enables customers to pay for meals by scanning their faces.
Customers place orders through a digital menu, and then use face scan as a payment
option.
Introduction 14
 Biometric identifier: eye recognition

Eye
Biometric recognition based on features
of the eye is compelling as the features
of our eyes have high uniqueness,
including the iris, retina and blood
vessel pattern in the whites of our eyes
(scleral vasculature recognition).
Systems using eye biometrics are
common for both identification and
verification. Applications ranges from
law enforcement and government
facilities to its recently integration in
mobile biometric authentication.

Introduction 15
 Biometric identifier: eye recognition

Eye - Retina
The first biometric eye scanning systems were retinal scanners
introduced in 1985. The retina is a thin tissue at the back of the
eye that has a pattern of capillaries that is unique. Not even
identical twins share a similar pattern, furthermore, the retina
remains unchanged from birth to death (if there are no eye
diseases). As the retina sits at the back of the eye and is not
directly visible, the image capturing cannot be done by a
conventional smartphone camera or similar device. The retinal
scan is instead performed by a beam of infrared light into the
eye which traces a standardized path on the retina detecting
the pattern of capillaries used for the authentication.

Introduction 16
 Biometric identifier: retina recognition

Retinal scanning has been used for identification in high

security settings by law enforcement agencies and
government organizations, e.g. FBI and CIA. Retinal
scanning is also used in medical diagnostic applications.
Although retinal scans provide a high degree of
security, the technology have many disadvantages
that has resulted in limited commercial use:
▪ Difficult to acquire useful image
▪ Subject must be very close to the camera optic
▪ Lengthy enrollment
▪ Requires specialized equipment
▪ Expensive scanners

Introduction 17
 Biometric identifier: iris recognition

Eye - Iris
The iris is the colored part of the eye, which controls the inflow of light to the eye. Specific
features of the iris, e.g. crypts, filaments, frills, collarette, furrows, Wolffin nodules, and
nevis, form patterns that are unique to a person. These features are captured using
CCD/CMOS image sensors and can be highlighted by near-infrared light (NIR: 700–900 nm).

Introduction 18
 Biometric identifier: iris recognition

The recognition accuracy is improved when the eye is illuminated with NIR
light. People with "dark eyes“ reveal less visible texture in the visible
wavelength band, while they appear richly structured in the NIR band.
Iris recognition is a relatively young technology, firstly patented in 1994. Iris
recognition was first used for identification systems in law enforcement and
government settings, and the recent advancements on camera technologies
and algorithms of iris recognition have allowed its use into mobile devices.
Some commercial iris-recognition systems can be easily fooled by high-quality photograph of a face,
making these scanners unsuitable for unsupervised applications, i.e door access-control systems.
Anti spoofing methods to provide some defense against the use of fake eyes and irises include:
- Changing the intensity of lighting during the recognition to verify pupillary reflex and to record the
iris image at several different pupil diameters
- Analyzing the 2D spatial frequency spectrum of iris image to find peaks caused by printer dither
patterns, typical of commercially available fake-iris contact lenses
- Using spectral analysis instead of monochromatic cameras to distinguish live iris tissue
- Observing the characteristic natural movement of eyeballs
- Testing for retinal retroreflection (red-eye effect)
Introduction 19
 Biometric identifier: iris recognition

The device is equipped with a dedicated iris camera, which utilizes a special image filter to receive and recognize the
reflected images of the irises with a red IR LED light.
The IR light allows better performance in iris scanning than traditional visible (or RGB) images, which can be affected by iris
color or ambient light, infrared images display clear patterns and have low light reflection.
The device received the highest International Electrotechnical Commission (photo biological stability) certification level. In
addition, the device will automatically switch off if it detects that the human eye is too close or exposed too long to the IR
LED sensor.

Introduction 20
 Biometric traits: iris recognition

Pros
▪ Very high accuracy – Little risk for false authentication
▪ Contactless – Makes it suitable for identification
Cons
▪ Speed – Iris recognition is not as fast as fingerprint recognition
▪ Processing requirements/power consumption – The processing power needed for
running the algorithms is higher than for fingerprint recognition, typically by a factor of
ten. Iris recognition also typically requires more data space to save the template than
fingerprint recognition.
▪ Spoofability – Iris recognition technology may be vulnerable to counterfeit techniques
and can have low security against a variety of spoofing attacks.

Introduction 21
 Biometric traits: scleral vein recognition

Eye – Scleral vein (eyeprint)

Scleral vein recognition is an
emerging biometric technology
that has been integrated recently in
some smartphones. The technology
uses the regular CMOS (or CCD)
image sensor in the device to
capture an image of the eyes.

A template based on the unique pattern of the scleral vasculature as well as other micro
features in and around the eyes is created from the image and the template. The primary
advantage of scleral vein recognition is that it requires no specialized hardware; a camera is
enough.

Introduction 22
 Case study - iris recognition at CERN

Access control in restricted area of CERN

The Challenge
CERN, located near Geneva on the Swiss-French
border, is recognized as the world’s leading
particle physics laboratory. The facility’s security
team needed improved access control to maintain
the safety of international scientists and staff,
irreplaceable research data and high-value assets.
The solution needed to be accurate, quick and
reliable – all while maintaining a database of more
than 10,000 registered people.
Requirements
▪ Manage a large database of scientist and staff
identities
▪ Control access to CERN’s underground
facilities
▪ Protect worker safety by limiting access to
hazardous areas
▪ Scalability to meet the needs of new facilities
Introduction 23
 Case study - iris recognition at CERN

The Solution
After completing a 2008 study of
available access control options,
CERN selected an iris-recognition
system from Iris ID Along with
greater speed and accuracy, the
iris-based system was judged to
be more hygienic practical and
capable of supporting a large-
scale application.
CERN had 55 iris readers installed at the main entry and access points to
certain work areas along the 27-kilometer circular Large Hadron Collider. The
LHC is the main underground particle collider hosting most of the
organization’s experiments.
The Iris ID system quickly moves 3,000 people daily. Identify authentication
requires less than a second. The process is safe, involving no contact, lasers
or bright lights.

Introduction 24
 Biometric traits: vein recognition

Palm/finger vein
Palm and fingers has a complex vascular pattern
that is unique to every person. Systems for vein
capture are based on infra-red light scanners.
Since the vein patterns lie under the skin, they
are almost impossible to replicate/spoof and
allows for highly secure authentication. The high
security levels and the contactless recognition
make vein recognition well suited for many
applications requiring very high security. What
limits the application areas are the size and cost
of the scanners. Identification takes considerable
time due to the high processing requirements, as
the vein patterns are very complex.
Introduction 25
 Biometric traits: hand geometry recognition

Hand and finger geometry

Features related to the human
hand (e.g., length of fingers)
are relatively invariant and
peculiar (although not very
distinctive) to an individual.
The image acquisition system
requires cooperation of the
subject to capture frontal and
and side view images of the palm flatly placed on a panel. The template storage
requirements of the hand are very small, which is attractive for bandwidth- and memory-
limited systems. Due to its limited distinctiveness, hand geometry-based systems are only
used for verification and not for identification. Finger geometry systems measure the
geometry of at most two fingers and may be preferred because of their compact size.
Introduction 26
 Multimodal systems

To increase the performance of biometric systems, it is possible to combine different

information for the decision making. Different alternatives are available such as combining
two different biometric systems (e.g., face and fingerprint), using two sensors of the same
modality (optical and sweeping fingerprint sensors), using two different algorithms given a
single capture, exploiting different representations of a single biometric modality (2D and 3D
face information) .
The combination of the decision results given by these multiple biometric sensors can be
realized from the easiest way based on a logical combination (conjunction) to more
complicated logic methods. Even if the global performance of multi-modal biometric systems
is improved, two main drawbacks may make this solution rarely used in our daily life:
▪ Expensive – multiple sensors are necessary
▪ Limited usability - Users may need to carry out multiple operations to be authenticated

Introduction 27
 Case study: ZKTeco 3-in-1 palm recognition technology

The 3-in-1 palm recognition process can be finished in 0.35 second.

Enhanced the recognition performance: the angle tolerance is as wide
as +/- 60 degrees (roll axis); the recognition distance can be up to 0.5
meters upon the palm-size. The anti-spoofing ability has reached a
new height of the industry, as the testing result shows, that all the
testing subjects, including fake HD photos, fake HD videos, and fake
palm models cannot stand any chance to crack the system.

https://zkteco.eu/news-center/news/access-control-palm-recognition-technology
Introduction 28
 Biometric identifier: DNA

DNA testing is highly accurate, considering that a person genome is

unique. The chance of 2 individuals sharing the same DNA profile is less
than one in a hundred billion, if 26 different DNA locations are studied.
• Intrinsically digital and does not change during a person life (or death)
• Sophisticated and expensive technology to conduct DNA test →
sample preparation, DNA amplification and data processing
• It requires a physical sample -> physically present person with
intrusive testing or biological samples in forensics.

Introduction 29
 Biometric traits: voice recognition

Voice recognition is unobtrusive and voice may be the only feasible biometric in applications
requiring person recognition over a telephone. Voice is not expected to be sufficiently
distinctive to permit identification of an individual from large databases of identities.
Moreover, a voice signal available for recognition is typically degraded in quality by the
microphone, communication channel, and digitizer characteristics. Voice is also affected by
factors such as a person’s health (e.g., cold), stress and emotional state. Besides, some
people seem to be extraordinarily skilled in mimicking others voice.

Introduction 30
 Biometric traits: voice recognition

There are two main approaches to voice authentication:

▪ Text-independent: Voice authentication is performed using any spoken passphrase or
other speech content.
▪ Text-dependent: The same passphrases are used in enrollment and for verification. This
means that a speaker cannot say anything he or she would like to authenticate, but will be
asked to speak a predetermined phrase. In static text-dependent voice authentication, the
same passphrase is used for every verification. In dynamic text-dependent voice
authentication, a randomized passphrase such as a number sequence is generated for the
user.

Voice authentication’s primary use case is hands-free mobile authentication. E.g. mobile
phones, automotive, virtual assistants (Google Home or Amazon’s Alexa), customer-support
calls.

Introduction 31
 Behavioural identifier: signature

The way a person signs his name is characteristic of the

individual. Signature recognition has been acceptable
in government, legal, and commercial transactions as a
method of verification. Signature is a behavioral
biometric that changes over time and is influenced by
physical and emotional conditions of the signatories.
Furthermore, professional forgers can reproduce
signatures of others to fool the unskilled eye.

Introduction 32
 Behavioural identifier: gait

Gait refers to the manner in which a person walks. The gait

parameters measured are typically both spatial-temporal (step
length, step width, walking speed) and kinematic (hip joint rotation,
mean joint angles of the hip/knee/ankle) counting for correlation
between step length and height. Accurate measurement of gait
parameters requires sophisticated equipment, as several video
cameras with floor-mounted load transducers, which makes secure
gait recognition a complicated and costly technology to implement.

Introduction 33
 Behavioural identifier: gesture

An alternative to gait recognition is gesture recognition,

where the handmade gestures are interpreted by
mathematical algorithms. Gesture recognition evolved
as a means for simplified human to computer
interactions and has become quite popular, e.g. in
computer games. Using gesture recognition for
authentication purposes is still far from being a secure
and spoof free authentication method, even if the main
advantage is the simple need of a camera.

Introduction 34
 Biometric system: system performance

The correct design and choice of a biometric system depends on system performance, social
acceptability and circumvention level.
The system performance is based on the following factors:
▪ Cost efficiency. How cost efficient is the biometric, a blended judgement for different
applications
▪ Processing speed. The speed of generating a template and matching it to one or many
stored templates, affects processing power required and hence also power consumption
– the higher speed the better
▪ Security. A blend of several factors, including spoofability and false acceptance rate
▪ Accuracy. How accurately the scanning devices can capture the biometric data in
different environments and how prone they are to disturbances

Introduction 35
 Biometric system: social acceptability

The social acceptability is evaluated by considering the following factors:

▪ Ease of use. How easy and intuitive it is to use for the user
▪ Popularity. How well known and how widely used the biometric technology is
▪ Acceptability. How supportive the society as a whole is of the technology
▪ Introduction year. The year that a biometric came into use

Introduction 36
 Acquisition of the biometrical trait: sensors

A sensor is a device that receives a signal or stimulus and responds with an electrical signal.
Sensors and their associated circuits are used to measure various physical properties such as
temperature, pressure, flow, position, light intensity, etc. These properties act as the stimulus
to the sensor, and the sensor output is conditioned and processed to provide the
corresponding measurement of the physical property.

Introduction 37
 Sensors

Sensors do not operate by themselves. They are generally part of a larger system consisting of
signal conditioners and various analog or digital signal processing circuits. Indeed, the full-
scale outputs of most sensors (passive or active) are relatively small voltages or currents, and
therefore their outputs must be properly conditioned before further analog or digital
processing can occur. Because of this, an entire class of circuits have evolved, generally
referred to as signal conditioning circuits.

Introduction 38
 Semiconductor industry

The semiconductor industry is based on the foundry model, which consists of semiconductor fabrication plants
(foundries) and integrated circuit design operations, which may belong to separate companies or subsidiaries.
Industries produce a variety of semiconductor components: in this course we’ll study some of solid-state
sensors and related integrated circuits, such as CCD cameras, CMOS cameras, capacitive sensors, and other
discrete components. Sensors, microprocessors and also the
analog components are fabricated by putting – that is, integrating
- on a single chip of a Silicon wafer the sensitive components with
the many transistors, from tens up to tens of billions →
integrated circuits (IC).

Introduction 39
 IC fabrication processes

In a simplified approach, the production

process for semiconductors, and in
particular ICs, consists of three distinct
steps:
➢ Design
➢ Fabrication
➢ Assembly
➢ Packaging
➢ Test

Integrated device manufacturers (IDMs), such as Intel or Samsung, SKHynix or Micron run (almost) all steps in-house. This has
been the first business model of the semiconductor industry but, starting in the 80s, it evolved due to the increasing
complexity and costs associated with design and fabrication of leading-edge ICs. Then, many companies specialized in single
production steps.
Introduction 40
 Scaling down of technologies

Miniaturization of electric components, first of all transistors, is the technological basis of the continuous improvement of
the chip performance over the decades, as depicted by Moore’s law. The number of transistors on integrated circuits
doubles approximately every two years → true exponential growth!

Dimension of single transistor

Introduction 41