1
Biometrics and Applications

Christophe CHARRIER1, Christophe ROSENBERGER1 and Amine NAIT-ALI2

1GREYC, Normandy University, University of Caen, ENSICAEN, CNRS, France

2LISSI, University of Paris-Est Créteil Val de Marne, France

Biometrics is a technology that is now common in our daily lives. It is notably used to secure access to smartphones or computers. This chapter aims to provide readers with an overview of this technology, its history and the solutions provided by research on societal and scientific issues.

1.1. Introduction

There are three generic ways to verify or determine an individual's identity: (1) what we know (PIN, password, etc.); (2) what we have (badge, smart card, etc.); and (3) what we are (fingerprint, face, etc.) or what we know how to do (keystroke dynamics, gait, etc.). Biometrics is concerned with this last set of approaches. Biometrics, and more precisely security biometrics, consists of verifying or identifying the identity of an individual based on their morphological characteristics (such as fingerprints), behavioral characteristics (such as voice) or biological characteristics (such as DNA).

The biometric features by which an individual's identity can be verified are called biometric modalities. Examples of some biometric modalities are shown in Figure 1.1. These modalities are based on the analysis of individual data, and are generally grouped into three categories: biological, behavioral and morphological biometrics. Biological biometrics is based on the analysis of biological data related to the individual (saliva, DNA, etc.). Behavioral biometrics concerns the analysis of an individual's behavior (gait, keyboard dynamics, etc.). Morphological biometrics relates to particular physical traits that are permanent and unique to any individual (fingerprints, face, etc.).

Figure 1.1. Examples of biometric modalities used to verify or determine the identity of an individual

Nowadays, the use of facial or fingerprint recognition has come to feel natural to many people, notably among the younger generations. Biometric technology is part of our everyday lives (used for border control, smartphones, e-payment, etc.). Figure 1.2 shows the spectacular evolution and market prospects of this technology. In an increasingly digital world, biometrics can be used to verify the identity of an individual using a digital service (social network or e-commerce). While fingerprints and facial or iris recognition are among the most well-known biometric modalities (notably due to their use in television series or movies), a very wide range of biometric data can be captured from an individual's body or from digital traces. An individual can be recognized in the physical and digital worlds using information from both spheres.

The use of this technology raises a number of questions: how new is this technology? How does a biometric system work? What are the main areas of current and future research? These questions will be addressed in the three main sections of this chapter: the history of biometrics (section 1.2), the technological foundations of biometrics (section 1.3) and the scientific issues and perspectives (section 1.4).

Figure 1.2. Evolution and perspectives of the biometrics market (source: Biometric System Market, October 2019)

1.2. History of biometrics

Biometrics may be as old as humanity itself. In essence, biometrics relates to a measurement that can be performed on living things, and in a security context, it refers to the recognition of individuals by their physical and/or behavioral characteristics. This property of recognition is primarily human based, and not dependent on technology. As humans, we recognize one another through aspects such as facial features, hands or gait; the human brain has the capacity to distinguish, compare and, consequently, recognize individuals. In reality, biometrics - as we now understand it - is simply a technological replication of what the human brain can do. Key aims include speed, reproducibility, precision and memorization of information for populations of theoretically infinite size (Nait-Ali and Fournier 2012).

From the literature, we find that biometrics began to be conceptualized several centuries BC, notably in the Babylonian civilization, where clay tablets used for trading purposes have been found to contain fingerprints. Similarly, fingerprinted seals appear to have been used in ancient China and ancient Egypt. It was not until the 14th century, however, that a Persian book, entitled Jaamehol-Tawarikh, mentioned the use of fingerprints for individual identification. Other later publications concerning the fingerprint and its characteristics include the work of G. Nehemiah (1684), M. Malpighi (1686), and a book published in 1788, in which the anatomist J. Mayer highlighted the unique nature of papillary traces.

It was only during the industrial revolution, notably in the mid-19th century, that the ability to clearly identify individuals became crucial, particularly due to an intensification of population mobility as a result of the development of commercial exchanges. The first true identification procedures were established in 1858, when William Herschel (working for the Indian Civil Service at the time) first used and included palm prints, then fingerprints, in the administrative files of employees (see Figure 1.3). Later, several medical scientists, anthropologists and statisticians, including Henry Faulds, Francis Galton and Juan Vucetich, developed their own studies of fingerprints. Vucetich was even responsible for the first instance of criminal identification using this technique, which took place in Argentina in 1892 (the Francisca Rojas case).

Figure 1.3. a) William James Herschel (1833-1917), and b) example of palm and finger prints (source: public domain)

A further turning point in biometrics occurred in the 1870s when Alphonse Bertillon, a French police officer, began to implement anthropometric techniques which came to be known as the Bertillon System, or "bertillonnage". Broadly speaking, this involved taking multiple measurements of the human body, including the face and hands. By combining these measurements with a photograph of the person and other physical descriptions (see Figure 1.4), Bertillon developed files which could be used to identify criminals and delinquents, even if they were disguised or using a false identity (see Figure 1.5). The first criminal identification using this technique in France occurred in 1902: Henri Léon Scheffer was identified by matching fingerprints taken from a crime scene with the information on his anthropological documents. At this time, the Bertillon system was used to a greater or lesser extent in many countries around the world.

Some 30 years later (1936), an ophthalmologist, Frank Burch, introduced the concept of identifying individuals by iris characteristics, although Burch did not develop this idea into an identification system. Biometrics as we now understand it began to take shape in the 1960s, drawing on technological advances in electronics, computing and data processing. The first semi-automatic facial recognition system was developed by the American Woodrow W. Bledsoe (Bledsoe and Chan 1965). The system consists of manually taking the coordinates of the characteristic points of the face from a photograph. These coordinates are then stored in a database and processed by computer by calculating distances with respect to reference points. In the same year, the first model of the acoustic speech signal was proposed by Gunnar Fan, in Sweden, laying the foundations for speech recognition. The first automatic biometric systems began to appear in the 1970s. Notable examples include a system for recognizing individuals by hand shape (1974), a system for extracting minutiae from fingerprints (FBI, 1975), a facial recognition system (Texas Instruments, 1976), a patent for a system for extracting signature characteristics for individual verification (1977), a patent for an individual verification system using 3D features of the hand (David Sidlauskas, 1985), a patent for the concept of recognizing individuals by the vascular network features at the back of the eye (Joseph Rice, 1995) and a patent for the concept of identifying individuals by characteristics of the iris (Leonard Flom and Aran Safir, 1986); the algorithm for this final system was later patented by John Daugman in 1994.

Figure 1.4. Plate taken from the Identification Anthropométrique journal (1893). a) Criminal types. b) Anthropometric file

Figure 1.5. Example of an anthropometric file using the Bertillon system (source: public domain)

The 1980s-1990s also saw an upsurge in activity with respect to facial recognition, notably with the application of principal component analysis (PCA) techniques by Kirby and Sirovich in 1988 (Kirby and Sirovich 1990), then the introduction of Eigenfaces by Turk and Pentland (1991). Turk and Pentland's paper was well received by the biometrics community, and has been cited over 18,500 times at the time of writing (2020). The authors demonstrated facial recognition using a limited number of parameters (compared to the number of pixels in a digital image), permitting the use of real-time...