|
" Improving the privacy, security, and performance of biometric systems "
Walter Jerome Scheirer
T. Boult
| Document Type
|
:
|
Latin Dissertation
|
| Language of Document
|
:
|
English
|
| Record Number
|
:
|
54470
|
| Doc. No
|
:
|
TL24424
|
| Call number
|
:
|
3357872
|
| Main Entry
|
:
|
Walter Jerome Scheirer
|
| Title & Author
|
:
|
Improving the privacy, security, and performance of biometric systems\ Walter Jerome Scheirer
|
| College
|
:
|
University of Colorado at Colorado Springs
|
| Date
|
:
|
2009
|
| Degree
|
:
|
Ph.D.
|
| student score
|
:
|
2009
|
| Page No
|
:
|
304
|
| Abstract
|
:
|
Biometric systems now represent a critical piece of the world's security infrastructure. Such systems are deployed at our borders, our workplaces, and even our amusement parks. Biometrics have found their way into military security checkpoints in Iraq and naval and coast guard operations around the globe. From the simplest time and attendance application, to the most sophisticated access control security installation, biometrics present a common set of benefits and problems for those that use and maintain such systems. This dissertation examines issues in three distinct, yet related components of biometric recognition systems: Privacy, Security, and Performance. Beginning with Privacy , a series of new attacks against existing template protection schemes is introduced, leading to a formulation of sound security requirements for revocable template technologies. Two novel secure revocable biotoken schemes are then introduced that conform to the aforementioned security requirements, including a base biotoken encoding method, and a bipartite biotoken method supporting secure data release. Using the biotoken technology as a basis, the Security of biometrics is considered through the development of a series of bio-cryptographic protocols, culminating in the development of a full biometric key infrastructure that enhances existing public key infrastructure. To address the threat of physical spoofing attacks, a low-cost spectrometer design is introduced that is capable of distinguishing between real live fingers and common artificial material. Regarding Performance , issues in long-range unconstrained face acquisition are discussed, followed by the introduction of a new facial feature detector and secure face recognition approach designed for this specific problem. Finally, a machine learning approach and a pure statistical approach are introduced for predicting the failure of biometric recognition systems. This research, taken as a whole, represents various embodiments of good and functional biometric systems that have yet to be realized in deployment.
|
| Subject
|
:
|
Applied sciences; Biometric systems; Fingerprint recognition; Face recognition; Computer vision; Biometrics; Computer science; 0984:Computer science
|
| Added Entry
|
:
|
T. Boult
|
| Added Entry
|
:
|
University of Colorado at Colorado Springs
|
| |