.png){kind=icon}
.png){kind=icon}
Return
IrisSPT: Cancelable Iris Template for Secure Authentication Based on Self-Parameterized Transform
.png){kind=icon}
PRE
AI
DOI：10.1109/TBIOM.2025.3557530.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Biometrics have received significant attention as a result of the increasingly widespread adoption of biometric authentication systems by government and commercial organizations as an alternative to traditional password or token authentication methods. Existing cancellable biometric techniques typically transform feature vectors into protection templates by generating random transformation functions. However, the transformation functions of such cancellable biometric systems are prone to leakage, which in turn may lead to security risks. For this purpose, a novel Self-Parameterized Transformation (SPT) method is proposed in this paper. Specifically, by generating the transformation parameters directly from the iris data, the method effectively prevents the attacker from establishing a recognizable mapping relationship between the original iris data and the generated template, which makes it extremely complicated to recover the original iris data. In addition, to ensure irreversibility, this method avoids storing the transformation parameters on the client or server side, thus effectively preventing the attacker from reconstructing the original iris data through reverse engineering. Although it is possible to attempt to recover the original data by brute-force cracking, this process is hardly feasible. To achieve unlinkability, this paper further employs an aggregation function, which effectively prevents correlation leakage between templates. Theoretical analysis shows that the proposed SPT method is adequately guaranteed in terms of irreversibility, revocability, unlinkability, and performance retaining criteria in cancellable biometrics. Benchmarking results show that the proposed SPT method improves at least 23.4% on the global measure (D-(sys)) compared to the existing schemes. Extensive experiments on CASIA-IrisV3-Interval, CASIA-IrisV4-Thousand and IIT Delhi Iris datasets demonstrate that the SPT approach has significant security advantages in biometric data protection and exhibits superior performance over state-of-the-art schemes. These results show that the proposed SPT method not only meets the stringent requirements in privacy protection, but also successfully realizes the balance between security and performance, and has high practical value.
Keywords:
Iris recognition
Biometrics
Protection
Security
Transforms
Servers
Feature extraction
Error correction codes
Vectors
Accuracy
Privacy protection
cancelable biometrics
non-invertible transform
self-parameterized transform
Journal
I
IF:
0
Papers:
67
Citations:
0