ICB

  2-5 June 2009 University of Sassari, Italy

  2009

 
 

The 3rd IAPR/IEEE International Conference on Biometrics

Home

Accommodations


Awards


Call for Papers


ICB 2011
call for proposals


Committee


Competitions


Conference Venue


Demo Submission


Keynote Speakers


Paper Submission


Program


Registration


Sponsors


Travel Information


Tutorials




ICB 2009 Tutorial Sessions
2 June 2009



ADAPTIVE BIOMETRIC SYSTEMS

Half-day tutorial: 4 hours (9am - 1pm)

Lecturers: Prof. Fabio Roli & Dr. Gian Luca Marcialis University of Cagliari, Italy, Dept. of Electrical and Electronic Eng.

Tutorial Contents and Objectives

Biometric recognition systems have been widely investigated in the last years. However, the increase of maturity of these systems has raised several problems. Among others, how dealing with intra-class variations, namely, variations of expressions, pose, lighting and other physiological or behavioural changes of the subject appearance, is still an open problem. These variations can be temporary or due to the intrinsic change of biometrics over the time ("aging" effect), and it is nearly impossible to represent them into the single template collected during the enrolment phase. Although multiple templates solutions have been proposed, they are mainly "supervised", that is, they require the human intervention. Recently, some novel approaches to the problem have been suggested, where the role of the human supervisor can be substantially reduced thanks to the ability of the biometric system of automatically adapting itself to subjects' intra-class variations. This is done by "template updating" approaches, which exploit data collected during the online system operation. The majority of these methods exploits concepts and methods coming from the "semi-supervised" learning theory. The goal of this tutorial is to focus on such novel approaches, and to survey the state-of-the--art in order to promote research on this important topic of adaptive biometric systems.



MEASURING THE GOODNESS OF CALIBRATION OF PROBABILISTIC BIOMETRIC RECOGNITION SCORES

Half-day tutorial: 4 hours (2.30 pm - 6.30 pm)

Lecturers: Niko Brümmer of Agnitio, South Africa and Prof. David A. van Leeuwen of TNO Human Factors and Radboud University in Nijmegen, Netherlands.

Tutorial Contents and Objectives

There are at least three ways to present the output of a biometric recognizer: (i) a single point estimate (hard decision) of the correct recognition hypothesis; (ii) a likelihood-based ranking of hypotheses (equivalently sortable recognition scores); (iii) a probability distribution over hypotheses. This tutorial is concerned with the third way. Provided that probabilistic outputs are well-calibrated, this is the most generally useful way of presenting recognizer outputs. A probabilistic output subsumes the other kinds of outputs and more generally allows the output to be optimally combined with other sources of (biometric, or non-biometric) information. In particular, probabilistic output allows the user to make optimal Bayes decisions-i.e. to do things like minimizing error-rate, minimizing cost, or optimizing utility. So if (iii) is better why is it not always used in preference to (i) or (ii)? It may be because it is perceived that (iii) is more difficult. One of the big reasons for this difficulty is that the goodness of probabilistic outputs cannot be measured by simply counting errors. This tutorial addresses that problem by motivating, analyzing and explaining the use of empirical cross-entropy to measure the goodness of probabilistic recognizer outputs. If researchers can measure (and agree on how to measure) the goodness of probabilistic recognizer outputs, then of course this is the first essential component towards improving the quality of probabilistic outputs.



BIOMETRICS: ETHICS, PRIVACY AND SECURITY

Full-day tutorial (9am - 1pm ; 2,30pm - 6.30pm)

Lecturers: Emilio Mordini, Centre for Science, Society and Citizenship, Rome, Terrance E. Boult and Walter J. Scheirer, Vision and Security Technology Lab, University of Colorado at Colorado Springs and Securics Inc.

Tutorial Contents and Objectives

This full day tutorial will cover ethics privacy and security of biometrics. The first part, will address the concept of identity and its ethical implications. The concept of personal identity is important from several perspectives. From a cultural perspective, the more the world converges, the more individual cultures wish to maintain their separate identities. From an individual perspective, the greater the population and the tendency to reduce people to stereotypes, the greater the desire to establish an individual identity. There is, however, another level where identity and the verification of identity, is becoming increasingly important in relation to all manner of transactions, from those related to mobility, to those related to legal, and political, rights and obligations, finally to financial and economical transactions. The intrusion of technology into these areas is not new, but their heightened visibility and ubiquity can create anxiety. This hold particularly true for biometrics. The second half of the tutorial will then overview security and privacy issues with traditional biometrics, introduce the Biometrics Dilemma, various threats it poses and a model for biometric DB risk highlighting the problem with standard large scale biometrics. The tutorial will explain why standard encryption does not solve the key problems, but also explore best practices in using standard encryption which can improve security. Moving to security, the tutorial will examine security system architectures, the role of authentication in such system and the standard architectures for authentication using biometrics. It will examine the advantages that biometrics bring, how biometrics can improve security and even privacy in such systems, and then discuss their weakness in both security and privacy. The tutorial will briefly discusses the the Nobel prize winning Economic theory of asymmetric information, Akerlof's market for lemons and Kerckhoffs' principles for security, and their implications for biometrics systems, especially large scale deployments. The last component the tutorial is an in-depth review the state of the art in what is sometimes called privacy preserving biometric technologies including biometric encryption, fuzzy vaults, fuzzy extractors, biometric hashing, and cancelabe biometrics. The tutorial will then walk through a security analysis of these technologies including the published attacks. The tutorial is intended to be relatively interactive with the opportunity discussion of some of the more subtle issues and a few "exercises" given out during the day with a discussion of the answers later in the day.



E-mail comments to webmaster at icb09.uniss.it
Last update : December 12, 2007