EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Improving Privacy in Iris Recognition Through Secure Matching of Protected Biometric Data

Osama Ouda

IEEE Access · 2026 · 14 · 86201-86220

DOI: 10.1109/access.2026.3698008 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Osama Ouda. Die Arbeit nennt 61 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Deanship of Graduate Studies and Scientific Research at Jouf University.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Privacy-preserving iris recognition remains a critical challenge in biometric authentication systems. While iris patterns offer exceptional discriminability for identity verification, their permanent and sensitive nature necessitates robust protection against template theft and privacy breaches. Traditional cancelable biometrics and cryptographic approaches often suffer from significant performance degradation or high computational overhead. This paper presents Indirect Protected Iris Matching via Parity–Magnitude Consistency, a new framework that re-evaluates biometric similarity assessment in privacy-preserving environments via indirect consistency testing. Rather than directly comparing biometric templates—which inevitably leaks sensitive structural information—the proposed method evaluates similarity indirectly through the collective response of a protected auxiliary structure to inter-sample discrepancies. The core innovation lies in transforming the matching problem from template comparison to statistical consistency testing on non-invertible, block-level transformations. We first establish a pure parity-based matching scheme that provides strong theoretical privacy guarantees but exhibits limitations on datasets with higher intra-class variability. To address this, we extend the framework to a hybrid parity–magnitude consistency approach that combines parity testing with Hamming weight verification, significantly enhancing discriminability while maintaining privacy. Comprehensive evaluation on the CASIA-IrisV3-Interval dataset demonstrates that the proposed hybrid method achieves an Equal Error Rate (EER) of 2.26% for block size $m=8$ , closely matching the unprotected baseline of 2.25%, while pure parity matching at the same block size yields an EER of 47.18%. The hybrid method maintains near-baseline accuracy for block sizes $m \le 16$ (EER $2.27- -2.36\%$ ) and degrades gracefully to 3.01% at $m=32$ , all while supporting renewability and being designed to resist inversion and cross-matching under the assumed attacker model. The method incurs modest additional storage overhead (a few bits per block for weight information) and adds minimal computational complexity (< 25% increase), establishing a practical balance between security, accuracy, and efficiency for real-world privacy-preserving iris authentication systems.

Förderung

  • Deanship of Graduate Studies and Scientific Research at Jouf University

Zitieren

APA

Ouda, O. (2026). Improving Privacy in Iris Recognition Through Secure Matching of Protected Biometric Data. IEEE Access, 14, 86201-86220. https://doi.org/10.1109/access.2026.3698008

Vancouver

Ouda O. Improving Privacy in Iris Recognition Through Secure Matching of Protected Biometric Data. IEEE Access. 2026;14:86201-86220. doi:10.1109/access.2026.3698008

BibTeX
@article{Ouda2026,
  author = {Ouda, Osama},
  title = {Improving Privacy in Iris Recognition Through Secure Matching of Protected Biometric Data},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {86201--86220},
  doi = {10.1109/access.2026.3698008}
}
Kostenlos starten →