Efficient Multi-Biometric Secure-Storage Scheme Based on Deep Learning and Crypto-Mapping Techniques

Cybersecurity has been one of the interesting research fields that attract researchers to investigate new approaches. One recent trends in this field is cancelable biometric template generation, which depends on storage a cipher (cancelable) instead original template. This trend ensures confidential and secure biometrics certain individual. paper presents multi-biometric system based deep fusion wavelet transformations. The part convolution (Conv.), transpose (Conv.Trans.), additional layers. In addition, deployed transformations are both integer transforms (IWT) discrete (DWT). Moreover, random kernel generation subsystem proposed work. method chaotic map modalities, including Baker modified logistic map. implemented four images, namely fingerprint, iris, face, palm images. Furthermore, it validated by comparison with other works literature. reveals shows superior performance regarding quality encryption confidentiality generated templates from input biometrics.