Design and Development of Template Transformation Techniques for Cancelable Biometric System

Abstract

Biometric recognition systems have become vital tool in recognizing an individual based on his/her unique biometric traits. Biometric templates derived out of the biometric traitscomprise unique features to identify an individual. Hence biometric traitsare being deployed in various access control and security applications currently. This has led to a raise in concerns over security and privacy of the biometric traits. These biometric templates, if compromised,cannot be revoked like passwords. Cancelable Biometric System is an effective biometric template securing scheme, which transforms the original biometric templates based on a user-specific key. Apart from template security, it also offers several other properties like non-invertibility, unlinkability, diversity and revocability to the biometric templates. With the revocability property, the biometric templates can be revoked or cancelled like a password whenever required. A new template can be generated to replace the cancelled template. Hence the system is called as Cancelable Biometric System , abbreviated as CBS henceforth. With the non-invertibility property, the template transformation process involved in CBS is strictly one-way. Hence the privacy of the biometrics is preserved. With the diversity property, a user can be enrolled to different applications with a corresponding cancelable template. With the unlinkability property, the diverse cancelable templates are uncorrelated to each other. Three different CBS have been proposed in this thesis work which can be applied inthree different scenarios, respectively. The first proposed Discrete Fourier Transform (DFT) +Random Projection (RP)-based CBS involves removing sparseness of the original biometric templates using DFT before applying cancelable transformation using Gaussian RP. The rationale behind the implication of DFT to the original biometric templateis to make it dense