Download files
Abstract
Cancellable fingerprint template has received increasing research interest since it was proposed. It not only possesses the ability to protect the original template from being retrieved by the attacker but is also able to revoke the old compromised template and reissue a new template. Many designs of cancellable template have been proposed. The performance of each of them increases each time a new work is proposed by showing lower error rates in verification. However, most of the proposed cancellable template designs not only deliver high Equal Error Rate (EER) when working with low-quality images but also show their weakness when dealing with Attack via Record Multiplicity (ARM). The ARM aims at bypassing the non-invertible transformation to acquire a user’s original fingerprint by gathering enough member equations of a system from multiple sources to solve for the variables in it. This research targets at developing a cancellable template design that delivers better result when working with low-quality images. Incorporating basic idea of clustering, the new method utilizes local structure to mitigate the negative influence of non-linear distortion. Achieving the low EER of 0%, 0.000017%, 0.02%, and 0.0598 % for FVC2002 DB1-3, and FVC2004 DB2, the proposed design shows incredible performance even working with low-quality images. Moreover, application of cancellable template in steganography in a secret image sharing scheme that can stand an ARM attack is also bought to reality: Secret image is embedded into fingerprint images distributedly. The fingerprint images are embedded into camouflage images that are stored with other dummy images on the cloud. Innovation in this work lies in the way that multiple impressions of the template fingerprint were used to protect the original features.