<p>Biometric systems play a crucial role in enhancing security for both identification and authentication tasks. Among them, finger-based biometrics remain one of the most reliable and widely adopted modalities, due to their ease of acquisition and the ability to capture multiple distinctive traits with a single sensor. In this work, we propose a multimodal biometric recognition framework that integrates fingerprint (FP) and finger vein (FV) modalities. The framework combines machine learning (ML) and deep learning (DL) techniques, incorporating a comprehensive data preprocessing and augmentation pipeline, transfer learning with pre-trained convolutional neural networks (CNNs) for feature extraction, and a dual-branch GraphSAGE-based inductive learning model to capture relational information among samples. The extracted embeddings are fused through an attention-driven feature-level fusion strategy, while an Out-of-Distribution (OOD) detection module is employed at the testing stage to filter anomalous inputs. Extensive experiments on three benchmark datasets demonstrate that the proposed system consistently achieves superior accuracy and outperforms state-of-the-art approaches, highlighting its robustness, adaptability, and effectiveness for secure biometric identification in real-world scenarios.</p>