<p>In digital identification technologies, fingerprint and iris-based biometric systems have found vast applications in user authentication. Hence, this study introduces a new hash-based technique and the concept of style transfer into the creation of biometric templates as an added measure to ensure the security of information and also add privacy to biometric systems. Iris images from both eyes and the left index finger fingerprint were used to generate cancelable biometric templates. In two sets, style transfer with deep learning was used: in the first set, the left eye was used as the content image and the left index finger as the style image; in the second set, the right eye provided as the content image and the left eye as the style image. A VGG16 was used for feature extraction. The extracted feature vectors were then concatenated and passed to a convolutional transfer layer to create a template image, which underwent histogram equalization. Different measures, such as the histogram, NPCR, SSIM, spectral distortion, and PSNR, were used for verification. To obtain the final hash value, the template image went through the SHA-256 method. Findings: A high NPCR of 99.90%, best SSIM of 0.547, lowest SD of 0.134, and greatest PSNR of 31.167 dB were obtained from the analysis of 200 samples, indicating that the method is successful in producing high-quality cancellable biometric templates. The proposed method effectively creates secure and cancellable biometric templates while preserving necessary discriminative features for accurate recognition. This approach provides a robust solution to biometric template protection, enhancing the security and integrity of biometric systems.</p>