Enhancing Security of Biometric Systems Using Deep Features of Hand Biometrics

Nowadays, the success of the different institutions depends on the security of their information against possible attacks. Because much of this information is related to persons' activities, so safeguarding their identities is essential for these successes. For this purpose, persons' identity recognition can be categorized as an effective method to enhance the security of information systems. To do this, most of these systems use biometric methods, which are very efficient compared to traditional methods. In this context, the present paper attempts to design an effective multimodal biometric system using hand biometrics. In this work, Finger- Knuckle-Print (FKP) and PaLMprint (PLM) biometrics are used. In addition, for high accuracy, two simple and effective deep learning based feature extraction techniques, PCANet and DCTNet, are integrated. The security analysis presented in this paper can help simplify the understanding of biometric systems and thus finding a robust solution to protect information systems. Our experimental results, using a synthetic multibiometric dataset of 165 persons, show that our proposed system has a high identification rate and a very low classification error compared to several existing methods.