GFT-GCN: Privacy-preserving 3D face mesh recognition combined with spectral diffusion

GFT-GCN combines graph-spectral learning with diffusion template protection for privacy-preserving 3D face mesh recognition. It uses GFT and GCN to extract spectral features and generates irreversible, renewable, and unlinkable templates through spectral diffusion; a lightweight client-server architecture keeps raw biometric data from leaving the client. On the BU-3DFE and FaceScape datasets, experiments show high recognition accuracy and strong resistance to reconstruction attacks.