aspg-2-3748-1791417275 20261007235435 American Scientific Publishing Group admin@americaspg.com American Scientific Publishing Group Journal of Cybersecurity and Information Management JCIM 2769-7851 2690-6775 2025 16 1 Louai A. Maghrabi Department of Software Engineering, College of Engineering, University of Business and Technology, Jeddah, Saudi Arabia In biometric applications, deepfake detection is a major field of research, as it is vital to certify the authenticity and integrity of biometric data. The manipulation of biometric information, like facial and fingerprint images, presents a critical attack on patient confidentiality and healthcare security. Deepfake is one of the manipulated digital media, for instance, an image or video of an individual can be substituted with a resemblance of another being. On the other hand, the growth of deepfake technology sets major attacks on biometric security by making hyper-realistic fake individualities that can deploy authentication methods. For deepfake recognition, a vital method in biometric applications utilizes a machine learning (ML) system, mainly deep learning (DL) that might study to differentiate amongst real and fake biometric data. In this manuscript, we present a Design of an Artificial Intelligence-Based Biometric Authentication System for Deepfake Detection with Patient Data Privacy Protection and Identity Verification (AIBADD-PDPPIV) algorithm. The main intention of the AIBADD-PDPPIV model is to deliver a secure and efficient biometric authentication approach that contributes to the advancement of privacy-preserving biometric security in healthcare systems. To accomplish this, the AIBADD-PDPPIV method employs an image preprocessing stage using the adaptive median filter (AMF) to reduce noise and enhance essential biometric features. For feature extraction, the vision transformer (ViT) model can be employed to capture intricate spatial dependencies in biometric images. Moreover, the multi‐head attention mechanism-based bidirectional gated recurrent unit (MA-BiGRU) model is exploited for deepfake detection and authentication processes. Eventually, the hyperparameter tuning process is accomplished through the pelican optimization algorithm (POA) to improve the detection performance of the MA-BiGRU model. To show the improved performance of AIBADD-PDPPIV model, a wide sort of simulations take place and the outcomes are inspected under numerous measures. The comparison study reported the betterment of AIBADD-PDPPIV system under various metrics. 2025 269 281 3748 https://creativecommons.org/licenses/by/4.0/ 10.54216/JCIM.160119 https://www.americaspg.com/journal/2/article/3748 https://www.americaspg.com/storage/21743618311.pdf