Global information security demands sophisticated solutions, with multimodal biometric systems enhancing recognition accuracy and overcoming single-mode system limitations. Unimodal biometric systems face challenges like forgery, user variability, and data security, which cannot be easily mitigated. Multimodal biometric systems offer enhanced security and robustness, addressing these limitations effectively. This project aims to develop a multimodal biometric identification system that fuses face and palmprint features at the feature level. It uses Convolutional Neural Networks to extract features from face and palmprint data. These feature vectors are integrated through feature-level fusion to improve identification effectiveness. The merged vector is then classified using a Support Vector Machine. The proposed multimodal system, evaluated using the real MULB dataset, significantly improves accuracy, recall, precision, equal error (ERR), false acceptance, and false rejection rates. The accuracy rates for individual palmprint and face systems are 97.30% and 98.97%, respectively, with EERs of 0.0162 and 0.0143. While the multimodal system achieves an improved accuracy rate of 99.28% and a low EER of 0.0095.