Multimodal LLM-Enhanced Biometric Authentication Framework: Integrating Dynamic Behavioral Patterns with Privacy-Preserving Analysis
Ravi Ray · Technix International Journal for Engineering Research · 2025
This article investigates the new use of Large Language Models (LLMs) alongside biometric authentication in addressing well-known problems in security, usability and privacy. The article framework takes advantage of LLMs’ pattern recognition skills to review and analyze multimodal biometric information from both physical and behavioral aspects. Turning traditional biometric data into forms that language models can use, the system delivers better authentication performance and maintains privacy by using differential privacy and federated learning. It is able to spot unusual actions, because it uses continuous risk checking based on current and historical context. Tests show that experimental evaluation outperforms regular GPS methods, especially in tough conditions and against advanced spoofing attacks. The strategy explained in the article improves security and solves important privacy concerns which could influence the way biometric technology is used in healthcare, financial and security