FaceLinkGen: A Re-evaluation of Identity Leakage in Privacy-Preserving Face Recognition and Face Anonymization Systems Using Simple Distillation
FaceLinkGen is a unified distillation attack framework targeting privacy-preserving face recognition (PPFR) and de-identification (De-ID) systems. It exploits the shared weakness that identity signals must remain available for utility, learning from paired protected and original images to regenerate faces or extract soft biometric attributes. This shows that resistance to fixed models does not prevent identity leakage.