Iris Template Protection using Discrete Logarithm

Abstract

Biometric authentication systems gained a huge public attention when compared to password authentication systems due to its direct connection with user identity. As Biometrics cannot be revoked or cancelled, several security and privacy issues will come if biometric templates are directly store into the database. To overcome this, there is a need to protect the biometric templates by applying transformations and at the same time accuracy won't be compromised. The transformed template is called cancellable template and it must satisfies all the requirements of Biometric Template Protection Schemes (BTPS) i.e., Diversity, unlinkability, accuracy, noninvertibility. The proposed method focused on generating cancellable iris templates by using discrete logarithm. By applying 1-D log Gabor filter on the iris images, iris codes were generated. Later a row vector is formed by appending next row to the previous one. Then the row vector is partitioned and converted into decimal vector. To achieve security or noninvertibility decimal vector is subjected to discrete logarithm over a prime field. To confirm the accuracy of the proposed approach, experiments are performed on CASIA-V 1.0 & CASIA-V3-Interval and achieved EER as 0.57% & 0.79%. Although the EER seems somewhat high, proposed approach is efficient in terms of security and noninvertible perspectives.