VAlue proposition
Face Recognition (FR) has become integral to identity management in many practical applications, from unlocking personal devices to facilitating law enforcement and accessing financial services. These systems process sensitive biometric data that, if compromised, can lead to privacy invasions, identity theft, and unauthorized surveillance. Unlike passwords, biometric data is immutable-once compromised, it cannot be changed, which elevates the need for robust security mechanisms to protect it. Such protections are also mandated by legal regulations on the acquisition, storage, and usage of biometric data, e.g., the European Union's General Data Protection Regulation (GDPR). FR systems in the wild consist of three entities: a probe face image, a feature extractor (i.e. a FR neural network), and a reference database of face features.
Description of Technology
This technology is a secure biometric recognition method in which all sensitive data remains encrypted throughout the entire recognition pipeline. The system encrypts an input biometric sample (such as a fingerprint, iris scan, or facial image) before sending it to a server or processor for feature extraction and biometric verification or search. By employing encryption techniques that allow computation directly on ciphertexts, the method ensures that the server never sees raw biometric data in unencrypted form. This helps protect individuals' privacy and secures biometric information against malicious attacks or unauthorized access. To efficiently handle the computational overhead introduced by encrypted operations, the system uses an encryption-compatible neural network that divides a biometric sample into manageable patches. Each patch is processed by a shallow, specialized network that runs in parallel. By keeping these networks relatively shallow, the overall multiplicative depth-and thus the latency associated with homomorphic encryption is greatly reduced. After local features are extracted from each patch, a lightweight fusion step aggregates the features to form a robust global representation. This design scales to various biometric modalities and resolutions, delivering improved speed and performance for end-to-end encrypted biometric recognition.
Benefits
Applications
IP Status
Patent Pending
LICENSING RIGHTS AVAILABLE
Full licensing rights available
Developer: Vishnu Boddeti and Wei Ao
Tech ID: TEC2025-0142
For more information about this technology, contact Raymond DeVito Ph. D. at devitora@msu.edu or 1(517)884-1658