الفهرس 
Chapter 1: IntroductionOnly 14 pages are availabe for public view
 |  | from | 75 |  |  |
| | | from | 75 | | |
Abstract
Nowadays, biometric system security enhancement is a significant subject that deserves
consideration. This is due to the dangers that face traditional common recognition systems,
which utilize Personal Identification Numbers (PINs) that can be easily stolen. If hacking
attempts succeed in getting access to the storage database of original templates, utilization of
original biometrics to access user services might result in the biometrics being lost,
permanently. We use cancellable biometric templates to protect the original biometrics from
being compromised in order to address this issue and avoid using them. Cancellable biometric
systems depend on the production of distorted or encrypted biometric copies that are used in
place of the original biometrics throughout the verification process. In order to create deformed
non-invertible cancellable templates that may be saved in the database, this study offers a
unique approach for user authentication using single and multiple biometrics. The proposed
approach depends on adaptive filtering. The original biometrics are either masked with patterns
created from adaptive filters or fed into the adaptive filter as input to get the cancellable
templates. In both cases, the adaptation and optimization algorithm of filter weights is well
exploited to get the required templates. Hence, adaptive optimization of filter weights is
exploited to yield cancellable templates. The performance of the adaptive filter is optimized to
yield the best cancellable templates from the security and privacy perspectives. For achieving
data compression in a multi-biometric situation, the suggested architecture starts with Discrete
Cosine Transform (DCT). The security level of the created deformed encrypted templates is
then increased by using Double Random Phase Encoding (DRPE). Finally, masking patterns for
biometric templates are created using an adaptive filter. The patterns are uncorrelated, which
increases protection against identity theft and facilitates more accurate identification. The
proposed cancellable biometric recognition framework performs well in simulations,
demonstrating a high Area under the Receiver Operating characteristic curve (AROC) of
99.9824% and a close-to-zero Equal Error Rate (EER). Other statistical assessment criteria have
also been taken into account to demonstrate the superiority of the suggested framework.