Privacy-Preserving Face Recognition Using Frequency Domain Differential Privacy

To overcome these limitations, a research team from China has proposed a new method for privacy-preserving face recognition. The proposed method utilizes differential privacy in the frequency domain, which offers a theoretical guarantee of privacy. The authors’ framework for face privacy protection based on differential privacy is fast, efficient, and allows for customizable privacy-preserving capability through the selection of a privacy budget. They also designed a learnable privacy budget allocation structure that protects privacy while minimizing accuracy loss in the differential privacy framework. In conclusion, the research team’s method offers a promising solution for privacy-preserving face recognition.