Fusion Strategies And Open-Set Recognition In Multimodal Biometric Authentication Systems: A Comprehensive Review

Multimodal biometric authentication systems have gained significant attention due to their ability to enhance reliability, accuracy, and resistance to spoofing by integrating complementary biometric traits. At the same time, open-set recognition (OSR) has emerged as a critical paradigm for real-world security applications, where previously unseen or unauthorized identities may appear during deployment. Despite substantial progress in unimodal OSR and multimodal fusion strategies independently, their integration remains limited. This paper presents a review of fusion strategies and open- set recognition in multimodal biometric authentication systems. We examine unimodal open-set methods and multimodal fusion techniques across sensor, feature, score, and decision levels. A comparative analysis is performed based on architecture, threshold design, scalability, and open-set awareness. The review reveals a structural gap between open-set biometric theory and practical multimodal fusion systems, particularly in threshold calibration, enrollment scaling, and fusion-level risk modeling. Additionally, an adaptive preset-based fusion framework is outlined as a potential approach to address these challenges. Finally, key research directions are identified to guide the development of unified open-set multimodal authentication frameworks for secure real-world deployment.