Matthew C. Valenti

This research area focuses on the design of secure, privacy-preserving biometric systems, with particular emphasis on biometric template protection, robustness, and system-level capacity. The work spans multiple biometric modalities—including face, iris, periocular, and multimodal biometrics—and integrates modern machine learning techniques such as deep hashing, generative adversarial networks, and neural network–based decoding. A central theme is the transformation of biometric features into secure representations that enable accurate authentication while preventing reconstruction, cross-matching, and information leakage. This includes learning error-tolerant biometric codes, embedding error-correcting structure into deep representations, and quantifying fundamental limits on biometric uniqueness and system capacity. The research also addresses practical deployment challenges, including cross-pose and cross-resolution recognition, cloud-based biometric services, and secure mobile authentication, providing a principled foundation for biometric systems that balance accuracy, scalability, and privacy.