Biocryptics: Towards Robust Biometric Public/Private Key Generation

We have developed a method for predicting variation in fingerprint samples. This enables the generation of a large database of public keys based on biometric data extracted from fingerprints of the users, without the need to use stored templates. A key pair consists of a private key, which is deleted immediately after enrollment, and a public key, which is stored in a database. The private key can be regenerated by processing a new sample of the fingerprint, thus needs only to exist when it is being used directly. To cope with variations in the sample, the key generation method allows minutiae points to change slightly without affecting the resulting key. Thus, the database does not only hold one public key for each user, but also keys that are likely (with high probability) to be generated if there is a small sample error in the future. This high probability is based on critical minutiae points.