Biometric authentication is becoming more crucial in contemporary society for secure identification and access control. This paper introduces a novel biometric authentication framework that leverages the complementary characteristics of finger knuckle and fingerprint features to improve the accuracy of the system. The key development involve use of Siamese neural networks with triplet loss function to combine deep, texture, and minutiae features extracted from the finger knuckle and fingerprint modalities effectively. The proposed system was rigorously tested on the PolyU Knuckle V1 and CASIA-Fingerprint V5 datasets, showing considerable performance enhancement compared to cutting-edge unimodal and multimodal biometric methods. The proposed methodology is justified its efficiency by gaining accuracy of 98.60%.