Published Paper Details:

Authentication is one of the most critical aspects of modern information security, ensuring that only legitimate users gain access to sensitive systems. Conventional single-mode authentication methods such as passwords or fingerprints are increasingly vulnerable to spoofing and hacking attempts. To overcome these limitations, multimodal biometric systems combine physiological and behavioral traits, thereby enhancing robustness. This paper presents a unified dual-mode authentication framework that integrates finger vein recognition with synchronized speech signals. The system also incorporates a two-step verification approach using KCP-DCNN with QR code validation to provide additional security. Ensemble-based classification is introduced using CNN, DBN, and Bi-LSTM tuned with SU-NBO optimization to improve accuracy and reduce error rates. Experimental results show superior performance with accuracy exceeding 97%, while maintaining low equal error rates across modalities. The proposed framework demonstrates resilience to spoofing, replay attacks, and environmental disturbances. The study highlights its applicability for high-security domains such as banking, healthcare, and IoT-enabled systems, paving the way for more reliable and secure authentication solutions.