Fingerprint liveness detection is an important aspect of biometric security systems to ensure that only live and genuine fingerprints are used for authentication.Fingerprint liveness detection algorithms are typically integrated into biometric devices, such as fingerprint scanners or sensors, to assess fingerprint characteristics and determine whether it is from a live finger or a spoof. Detecting liveness in fingerprints can be challenging as an artificial fingerprint can be crafted to mimic the texture, appearance, and even physiological properties of live fingerprints. The present study proposes a fingerprint liveness detection method using a convolutional neural network and its training models. Two CNN training models, i.e., Xception and InceptionV3, are used to obtain a better liveness detection system. The proposed model splits into two categories. For category one, the model is trained and tested through the ATVS-FFp-DB dataset and for category two, the model is trained and tested using the SocoFing dataset. Integrating the Xception and InceptionV3 deep learning models with the LPQ, LBP, and BISF texture descriptors yields exceptional results, achieving an accuracy rate approaching 99% across both Category-1 and Category-2 databases.

Biometrics, Fingerprint Liveness Detection, Security, CNN, LBP, LPQ and BSIF