Non-invertible geometrical transformation for fingerprint minutiae template protection

We propose a non-invertible geometrical transformation to protect minutiae-based fingerprint templates. This algorithm combines parameter-controlled linear and non-linear geometrical transformations to map an original minutiae-based template to a protected coordinate-based template in a highly non-invertible way. The protected template is easy to revoke and renew by changing the transformation parameters. Reference points based distance-preserved linear transformation is employed to preserve biometric performance, while cropping and superimposition based non-linear transformation is employed to achieve high non-invertibility which even permits all transformation parameters public in some cases. The high non-invertibiliy also enhances unlinkability among different templates for the same original fingerprint minutiae template against the key-inversion attack which undermines existing algorithms such as fuzzy commitment and fuzzy vault.