A KCP-DCNN-Based Two-Step Verification Multimodal Biometric Authentication System featuring QR Code Fabrication

Abstract Starting with for, need change Enhanced authentication performance, the concept of multi-biometrics authentication systems has emerged as a promising solution in today's digital era. In existing literature, numerous studies on multi-biometrics authentication have been carried out. However, such studies have proven their inefficiency in combining biometric and non-biometric for authentication and differentiating real and forged biometric data. Thus, an effective multimodal Biometric Authentication (BA) technique utilizing a Kernel Correlation Padding-based Deep Convolutional Neural Network (KCP-DCNN) is proposed in this paper. In the model, signature, fingerprint, and face modalities are combined. Primarily, the input images are preprocessed for image magnification utilizing the Radial Basis Function-centric Pixel Replication Technique (RBF-PRT) and augmentation utilizing Log Z-Score-centric Generative Adversarial Networks (LZS-GAN). Next, for FDivergenceAdaFactor-centric Snake Active Contour Model (FDAF-SACM) based contour extraction, Chaincode-centric minutia extraction, and Dlib's 68-centric facial point extraction, the magnified signature, magnified fingerprint, and augmented face images are utilized need combine with first part presented in the abstract. In this digital age, multi-biometric authentication systems have become a potential approach for improving authentication performance. Existing literature elaborates numerous studies on multi-biometrics authentication have been carried out. However, such studies have proven their inefficiency in combining biometric and non-biometric for authentication and differentiating real and forged biometric data. Thus, an effective multimodal Biometric Authentication (BA) technique utilizing a Kernel Correlation Padding-based Deep Convolutional Neural Network (KCP-DCNN) is proposed in this paper. In the model, signature, fingerprint, and face modalities are combined. Primarily, the input images are preprocessed for image magnification utilizing the Radial Basis Function-centric Pixel Replication Technique (RBF-PRT) and augmentation utilizing Log Z-Score-centric Generative Adversarial Networks (LZS-GAN). Next, for FDivergence AdaFactor-centric Snake Active Contour Model (FDAF-SACM) based contour extraction, Chaincode-centric minutia extraction, and Dlib's 68-centric facial point extraction, the magnified signature, magnified fingerprint, and augmented face images are utilized. Proposed technique augmented its precision, recall, and F-measure1.88%, 2.47%, and 1.19% than the prevailing CNN.Then, for efficient classification utilizing KCP-DCNN, significant features are extracted. If the classification output is real, then the user is authenticated after the verification of the Quick Response (QR) code generated utilizing the extracted points. The user identity is recognized with 98.181% accuracy by the developed model. Thus, the authentication rate of the Multimodal Biometric (MB) system is increased 98.8% accuracywhat percentage? by the proposed system. move this first part of the abstract.Then, for efficient classification utilizing KCP-DCNN, significant features are extracted. If the classification output is real, then the user is authenticated after the verification of the Quick Response (QR) code generated utilizing the extracted points. Thus, the authentication rate of the Multimodal Biometric (MB) system is increased by the proposed system.