In today’s digital world, biometric technologies play a pivotal role in ensuring security and access control. As the global demand for biometric systems grows, integrating multiple features becomes imperative. Getting useful information out of specific biometric modalities is the key to making these kinds of systems resilient. It is widely believed that multimodal biometric security systems are safer and more accurate than unimodal ones since they use a combination of features. To address the challenges posed by potential attacks using stolen or compromised biometric data, we present an advanced approach: Multimodal face, finger print and iris Biometric Fusion Using Hybrid Deep Learning for Enhanced Person Identification (MFFI-HDL-EPI). First, we employ using Deep Neural Network Algorithm (DNN) named Graph Convolutional Neural Network (GCN) that excels at capturing complex relationships in graph-structured data, making them suitable for multimodal feature fusion. Next, we employ Improved Group Search Optimization (IGSO) that dynamically adapts search strategies and parameter adjustments, leading to faster convergence and improved feature optimization. Our proposed Multimodal Fusion approach combines face, fingerprint, and iris biometric features. By leveraging GCNs and IGSO, we enhance person identification accuracy. For minimizing error metrics and improving prediction accuracy, DBN-SVM Meta-Classifier is presented. It uses DBN features as input to a Support Vector Machine (SVM) for final decision-making. Performance evaluation metrics include F-measure, recall, precision and accuracy. Additionally, we assess error metrics such as GAR (Genuine Acceptance Rate), FRR (False Rejection Rate), FAR (False Alarm Rate), and EER (Equal Error Rate). In summary, MFFI-HDL-EPI represents a significant advancement in multimodal biometric fusion, offering heightened security and accuracy for person identification.