With the fast growth of modern technology, biometrics has become a common method of verifying a person’s identity. However, a major challenge in biometric authentication is ensuring template security, which is essential to protect user privacy. To enhance this security, the original feature-based template is transformed into a pseudo-template before use. In this paper, we first extract discriminative features from the biometric trait. After feature extraction, features are then flattened and converted into secure binary templates using Physical Unclonable Function(PUF) scheme coupled with a Addition Rotation XOR (ARX) mechanism, ensuring both revocability and non-invertibility. The final fused template forms a cancelable biometric representation that protects against template compromise while preserving recognition performance. The proposed approach is evaluated on the Children Multimodal Biometric Database (CMBD), demonstrating strong privacy protection with high precision. The experimental results show that the proposed scheme achieves an Equal Error Rate (EER) of 0.014%.