الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 150 |  |  |
| | | from | 150 | | |
Abstract
Nowadays, biometric identification systems provide a reliable solution for
identity verification that allows restrictive access to nuclear facilities by
unauthorized personnel. Biometric identification systems offer enhanced accuracy,
improved accountability, and a reduction in opportunities for misuse, compared to
other traditional identification systems that relay on passwords, ID cards, or
personal identification numbers (PINs). However, biometrics also have some
specific security/privacy issues. For example, biometrics are vulnerable to be
captured easily without the user’s consent. Also, if a user biometric is exposed once,
it is compromised forever, because it is always associated with the user and cannot
be replaced or cancelled.
Moreover, biometric-based authentication methods depend on the same
biometrics. So, if a biometric template is compromised in an application, then the
same method can be used to access other applications at which the same biometric is
used. Furthermore, while relative robustness over time is an advantage for
biometrics, it can also be a massive challenge from the privacy point of view, when
the biometric needs to be changed. So, biometrics definitely are sensitive data, and
therefore they should be properly secured, because they may be misused by any
attackers. Therefore, integrating a high security solution into the biometric
recognition system to protect personal information during storage, and transmission
is a necessity.
In spite of proposing several methods for preserving the privacy and security
of the individual biometric data, this comes at the expense of decreasing the
performance accuracy in comparison to traditional unprotected biometric
recognition systems. This motivates us to design a new lightweight, robust and
secure technique for enhancing the security of biometric recognition systems
without affecting the performance accuracy.
This thesis presents a novel model for securing iris recognition systems by
using a combination of cancellable IrisCode generation based on salting approach
and iris cryptosystem based on elliptic curve cryptography. In this model, a- IV -
cancellable IrisCode is generated by combining the original binary IrisCode with a
totally artificial cover pattern (synthetic pattern) in a piel-wise manner using XOR
operation.
Once the non-invertible transformed version is obtained from the enrolled
IrisCode, the original biometric pattern is discarded. The matching stage is made
primarily on the resultant distorted cancellable iris pattern. Then, the cancellable
IrisCode is divided into binary blocks. Each binary block is mapped to a point on
the elliptic curve, and then encrypted based on elliptic curve encryption, generating
a protected IrisCode. Hence, the protected IrisCode can be stored and transmitted
more securely.
This model guarantees a high degree of privacy/security protection without
affecting the performance accuracy compared to the unprotected traditional iris
biometric system. The proposed technique has been analyzed in the absence and
presence of different types of noises, and then compared with the traditional
unprotected iris recognition technique. Robustness to scaling effects as well as
security of the proposed technique has been evaluated. Furthermore, a comparative
study of the proposed technique with the existing iris protection techniques is
presented and discussed.