The practice of using anatomical traits to determine the identity of an individual dates back to the late 19th century when Alphonse Bertillon advocated a personal identification system based on a set of anthropometric measurements. But the Bertillon system lacked automation and was cumbersome and tedious to use. Not surprisingly, it was abandoned in favor of a simpler and more accurate approach involving fingerprint comparison by human experts, which was made possible by the pioneering works of Faulds, Galton, Henry, and Herschel. Though fingerprint identification systems were adopted by several law enforcement agencies including the Scotland Yard at the beginning of the 20th century, it was not until 1963 that the first scientific paper on automated fingerprint matching was published by Trauring in the journal Nature. Trauring's work laid the foundation for modern day biometric recognition systems and was followed by the development of automated systems for matching other anatomical and behavioral traits such as voice (Pruzansky, 1963), face (Bledsoe, 1964), signature (Mauceri, 1965), hand geometry (Jacoby et al., 1972) and iris (Daugman, 1992). It is instructive to reflect on what progress has been made in biometric recognition over the past 50 years since Trauring's landmark paper. Significant progress has indeed been achieved making it possible to accurately recognize individuals based on biometric trait(s) (e.g., fingerprint, face, iris, or voice) acquired in controlled acquisition environments with user cooperation. While these developments have enabled a wide variety of identification applications ranging from personal laptop access to national civil registry systems, a number of thorny issues continue to inhibit the potential of biometric systems. One of the main unsolved issues is the problem of processing poor quality biometric data, possibly acquired from uncooperative users in unconstrained environments. Moreover, issues related to the security and privacy of the biometric data itself, robustness of the system to spoofing and obfuscation, and uniqueness and persistence of biometric traits need to be systematically studied. Unlocking the potential of biometrics through fundamental research in the context of these larger systemic issues will not only lead to widespread adoption of this promising technology, but will also result in user acceptance and societal good.

Anil K. Jain is a University Distinguished Professor in the Departments of Computer Science & Engineering, and Electrical & Computer Engineering at Michigan State University. He received a B.Tech. degree from IIT, Kanpur (1969) and M.S. and Ph.D. degrees from Ohio State University in 1970 and 1973, respectively. His research interests include pattern recognition, computer vision and biometric recognition. His articles on biometrics have appeared in Scientific American, Nature, IEEE Spectrum, Comm. ACM, IEEE Computer, Proc. IEEE, Encarta, Scholarpedia, and MIT Technology Review. He has received a number of awards, including Guggenheim fellowship, Humboldt Research award, Fulbright fellowship, IEEE Computer Society Technical Achievement award (2003), IEEE W. Wallace McDowell award (2007), IAPR King-Sun Fu Prize (2008), and IEEE ICDM 2008 Research Contribution Award for contributions to pattern recognition and biometrics. He also received the best paper awards from the IEEE Trans. Neural Networks (1996) and the Pattern Recognition journal (1987, 1991, 2005). He served as the Editor-in-Chief of the IEEE Trans. Pattern Analysis and Machine Intelligence (1991-1994). He is a Fellow of the ACM, IEEE, AAAS, IAPR and SPIE. He has been listed among the "18 Indian Minds Who are Doing Cutting Edge Work" in the fields of science and technology. Holder of six patents in the area of fingerprints (transferred to IBM in 1999), he is the author of several books, including Introduction to Biometrics (2011), Handbook of Biometrics (2007), Handbook of Multibiometrics (2006), Handbook of Face Recognition (first edition: 2005; second edition 2011), Handbook of Fingerprint Recognition (first edition: 2003, second edition: 2009) (received the PSP award from the Association of American Publishers), Markov Random Fields: Theory and Applications (1993), and Algorithms For Clustering Data (1988). ISI has designated him as a highly cited researcher (his h-index is 129). According to CiteSeer, his book, Algorithms for Clustering Data is ranked # 75 in the Most Cited Articles in Computer Science (over all times) and his paper Data Clustering: A Review (ACM Computing Surveys, 1999) is consistently ranked in the Top 10 Most Popular Magazine and Computing Survey Articles Downloaded. He is serving as a member of the National Academies panel on Information Technology and previously served on panels on Whither Biometrics and Improvised Explosive Devices (IED). He also served as a member of the Defense Science Board.

The links between forensic science and biometry go back to the early days of forensic science at the start of the 20th century. With the advent of computer science and automation in the 70s, biometric systems have provided major improvements in crime detection and deterrence. Indeed AFIS technology has drastically increased the efficiency and throughput of identification services at national and international levels. However biometric systems are essentially still used only as sorting devices, allowing a rapid and efficient search through millions of records. Biometric systems remain providers of leads (names) for further forensic evaluation in the same way police investigators are providing investigative leads, but at present they play no role in the subsequent forensic decision making process. For example, the decision of identification of a fingermark recovered from a crime scene to a given individual will be reached by the fingerprint examiner regardless as to whether or not the individual had been short-listed thanks to a biometric system.

In the last ten years, iris, face, ear recognition systems have improved to a stage that their application under unsupervised (uncontrolled) conditions, as it is typically the case in forensic science, is feasible. The scope of application of such technologies to law enforcement identification is rapidly expanding both in the early investigative phase, as an intelligence tool, or post criminal activity as an evidence-gathering tool. Alongside with this technological push, forensic science is under scrutiny. Traditional forensic fields (such as fingerprints, handwriting, face recognition) are under fierce criticisms for not being underpinned by strong systematic and structured studies. The expertise has been traditionally left to the opinion of the experts who developed their identification skills through training and experience, with technology having almost no role to play in that process. This changing landscape shapes new ways working between forensic science and biometric systems. This presentation will try to delineate them. Three case studies will help to explore the future opportunities and challenges facing the forensic and biometric community. These case studies will show:

1. How the advent in AFIS technology suggests new ways of operations for experts and operational units, including light-out identifications on fingermarks;
2. How face biometric systems (such as face recognition) can be deployed as an efficient investigative tool in law enforcement agencies;
3. How biometric systems can help experts to assign weights of evidence to comparisons arising from identification transactions. Particular attention will be given to the challenges posed by an operational deployment of such tools in forensic science.

Christophe Champod received his M.Sc. and Ph.D. (summa cum laude) both in Forensic Science, from the University of Lausanne, in 1990 and 1995 respectively. He remained in academia until holding the position of assistant professor in forensic science. From 1999 to 2003, he led the Interpretation Research Group of the Forensic Science Service (UK), before taking a professorship position at the School of Criminal Sciences (ESC) / Institute of Forensic Science (IPS) of the University of Lausanne. He is in charge of education and research on identification methods (detection and identification). He is member of the International Association for Identification and was elected in 2004 member of the FBI-sponsored SWGFAST. His research is devoted to the statistical evaluation of forensic identification techniques. The value of fingerprint evidence is at the core of his interests.

Being able to identify an individual or to confirm a claimed identity has been an essential activity of many existing and past societies. With the globalization of our modern world, this activity has become more and more challenging: many activities which used to occur at local level (village) between people knowing each other (local community) now occur at national or international level, involving huge populations from anywhere on the planet. This is making the need to identify/authenticate people more and important. This is also making this task more and more difficult, as a very large number of identifications (millions per day) are to be performed on very large populations (hundreds of millions of people). Biometric techniques is a unique and efficient tool for those who face these identification/authentication needs, and many organizations rely on large biometric systems to identify people: Governements, to know their population and deliver them citizen or social rights (such as voting, or social benefits) or to protect their borders. Law enforcement agencies, to identify dangerous individuals and to solve crimes. Corporate sector, to protect their assets (physical access control) or to secure transactions between consumers.

Designing such a large scale biometric system is a challenging problem combining the need for powerful algorithms with the complexity of building a scalable system and the challenge of its operational deployment in the field. Very large implementations exist today, with several systems around the world hosting more than a hundred million people and able to process over a million identification searches per day. The largest of those systems today is the UID deployment in India, with over 275M people enrolled and deduplicated. In this talk we will go through the challenges associated to the design and implementation of such systems and try to understand the gap between the original idea of a new algorithm method to its implementation deployment in the real world, going through the challenges of accuracy&performance, scalability and extrapolation, auto adaptivity and robustness to errors, usability issues.

Jean-Christophe is the VP, Research & Technology for the Identification division of Morpho. He has been leading Morpho Biometric research activity for more than 12 years and has been involved in all of Morpho's breakthroughs in biometrics of the past 18 years. His achievements includes the development of biometric algorithms and of innovative biometric sensors, the implementation of biometric algorithms in smart cards or the design of scalable architecture for very large biometric systems such as the FBI or UID in India. He holds more than 13 patents in the field of biometrics and is a senior member of IEEE and a member of the International Association for Identification.