Modeling IrisCode and its variants as convex polyhedral cones and its security implications.

Area of Science:

• Biometrics
• Computer Science
• Cybersecurity

Background:

• IrisCode, developed by Daugman in 1993, is a foundational iris recognition algorithm.
• Over 100 million individuals have been enrolled using IrisCode, influencing numerous biometric identification and template protection methods.
• Understanding IrisCode's structure is crucial for assessing current biometric security.

Purpose of the Study:

• To analyze the geometric properties of IrisCode templates.
• To develop a method for computing the central ray of IrisCode templates.
• To evaluate the security implications of this geometric representation for biometric template protection.

Main Methods:

• Representing IrisCode templates as convex polyhedral cones in hyperspace.
• Deriving a simple algorithm to compute the central ray of these cones, independent of original biometric data.
• Testing the central ray computation for matching across different biometric templates (iris, palmprint) and assessing the security of protected templates.

Main Results:

• IrisCode templates and their variants form convex polyhedral cones.
• A simple algorithm can compute the central ray, representing the original biometric signal.
• Biometric templates from different methods can be matched using their central rays.
• Templates protected by IrisCode-extended methods are vulnerable to being broken.

Conclusions:

• Convex polyhedral cone templates derived from IrisCode are not inherently secure without rigorous security analysis.
• The simplicity of the central ray computation method poses a significant risk, enabling unauthorized access by individuals with limited technical expertise.
• This research highlights potential security flaws in widely used biometric template protection schemes.