Cancelable Fingerprint Recognition based on Encrypted Convolution Kernel in Different Domains

Abstract

Peoples' biometrics, such as fingerprints, are unique , as a result it can be used in many evidence security requests, such as employees' registration gate, crime investigation, and revealing smart phones. The security of fingerprints is very critical in protecting the peoples' identity. In this research, a cancelable fingerprint recognition system is proposed. The proposed system is based on comprising four biometrics in a unified biometric template for each person using Discrete Cosine Transform (DCT) compression. This unified biometric template is encrypted with different convolution kernels produced by chaotic Baker map in different domains. The Integer Wavelet Transform (IWT) and the Discrete Wavelet Transform (DWT) are used to create different transformations of the fingerprint. In case a transformed fingerprint is compromised, the biometric fingerprint transformation is reorganized with another transformation. A comparative study between different transform-domains in the occurrence of attacks shows the authority of encryption in the DWT domain with different keys. Both Equal Error Rate (EER) and Area under Receiver Operating Characteristic (AROC) curve are used for performance evaluation revealing high performance of the proposed system.

Process of Fingerprint Authentication using Cancelable Biohashed Template

“…Model's precision is number of fingerprint that were correctly authenticated, divided by all fingerprints that the model picked. [20] 0.19% Partial local structure by Kho et al [9] 0.01% fractal coding and fourier mellin transform by abdullahi et al [1] 0.364% one permutation Hashing by Li et al [14] 0.19% Integer wavelet transform by Hashad et al [6] 0% proposed 3.13%…”