Virginia Tech® home

GenA(eye)ris: Generating Synthetic Iris Biometrics for Presentation Attack Security and Security for Generative Biometric Models

Researchers from Virginia Tech, George Mason University

Researchers aim to increase the security of iris biometrics by using generative AI to boost the performance of presentation attack detection (PAD) models and to safeguard generative models’ training datasets from leaking identifiers.

They’ll also explore the design of multi-modal biometrics extracted from the eye region to enhance accuracy, robustness, and security in identity verification within mixed-reality environments. 

Funded by the CCI Hub

Rationale

Iris biometrics are considered one of the most accurate and reliable biometric modalities, but the technology is vulnerable to presentation attacks (PAs).

A PA involves the use of an artifact that mimics the identity of a legitimate user, using physical and digital spoofing methods such as contact lenses, a high-resolution printout, or a replica created with an advanced 3D printer.

Projected Outcomes

Researchers will improve the robustness of PAD models against evolving spoofing techniques and utilize multi-modal biometric authentication to implement rigorous privacy measures to protect generative systems against data leakage and presentation attacks.

Results will be used for external grant proposals to advance the security of iris biometrics via generative AI and  to identify new research challenges, such as designing multi-modal biometric systems integrating iris patterns, periocular features, and gaze data.