Practical Biometric Authentication with Template Protection

We show the feasibility of template protecting biometric authentication systems. In particular, we apply template protection schemes to fingerprint data. Therefore we first make a fixed length representation of the fingerprint data by applying Gabor filtering. Next we introduce the reliable components scheme. In order to make a binary representation of the fingerprint images we extract and then quantize during the enrollment phase the reliable components with the highest signal to noise ratio. Finally, error correction coding is applied to the binary representation. It is shown that the scheme achieves an EER of approximately 4.2% with secret length of 40 bits in experiments