Cancellable Biometrics: A Privacy-Preserving and Adaptive Framework for Next-Generation Secure Authentication

Abstract

The proliferation of biometric authentication technology has raised issues relating to template security, privacy, irrevocability, and cross-application link ability. While there exist cancellable biometric techniques which offer template protection by employing non-invertible transformation techniques, most are characterized by low recognition performance, lack of flexibility across various biometric modalities, and susceptibility to advanced template attacks. In order to address the aforementioned shortcomings, this paper presents a new Adaptive Cancellable Biometric Framework (ACBF) which employs adaptive non-invertible transformation, modality aware template creation, and dynamic revocation techniques to offer secure, privacy enhancing, and revocable biometric authentication. In contrast to other cancellable biometrics which use predetermined transformation functions, the ACBF employs dynamic generation of cancellable templates for specific applications. The proposed approach is tested through the use of publicly accessible fingerprint, face and multimodal biometric data sets. The experimental results show that recognition accuracy can reach 98.2% for fingerprint-based authentication and 97.9% for face-based authentication with minimal loss of performance when compared to traditional non-cancellable biometric systems. In addition to this, recognition accuracy increases to 99.1%, while equal error rate decreases to 0.9%, because of multimodal fusion. Revocation tests have revealed that the value of correlation coefficient between original and reissued templates is less than 0.15, proving high revocability and cross application unlinkability.