Fusion of fingerprint presentation attacks detection and matching: a real approach from the LivDet perspective

The liveness detection ability is explicitly required for current personal verification systems in many security applications. As a matter of fact, the project of any biometric verification system cannot ignore the vulnerability to spoofing or presentation attacks (PAs), which must be addressed by effective countermeasures from the beginning of the design process. However, despite significant improvements, especially by adopting deep learning approaches to fingerprint Presentation Attack Detectors (PADs), current research did not state much about their effectiveness when embedded in fingerprint verification systems. We believe that the lack of works is explained by the lack of instruments to investigate the problem, that is, modelling the cause-effect relationships when two systems (spoof detection and matching) with non-zero error rates are integrated. To solve this lack of investigations in the literature, we present in this PhD thesis a novel performance simulation model based on the probabilistic relationships between the Receiver Operating Characteristics (ROC) of the two systems when implemented sequentially. As a matter of fact, this is the most straightforward, flexible, and widespread approach. We carry out simulations on the PAD algorithms’ ROCs submitted to the editions of LivDet 2017-2019, the NIST Bozorth3, and the top-level VeriFinger 12.0 matchers. With the help of this simulator, the overall system performance can be predicted before actual implementation, thus simplifying the process of setting the best trade-off among error rates. In the second part of this thesis, we exploit this model to define a practical evaluation criterion to assess whether operational points of the PAD exist that do not alter the expected or previous performance given by the verification system alone. Experimental simulations coupled with the theoretical expectations confirm that this trade-off allows a complete view of the sequential embedding potentials worthy of being extended to other integration approaches.