Defender of privacy and fairness: Tiny but reversible generative model via mutually collaborative knowledge distillation

Sissi Xiaoxiao Wu, Zehong Huang, Zhicong Liang, Lin Gu, Tatsuya Harada, Zheng Li, Yingying Zhu

Abstract: Highlights•We propose a effectively knowledge distillation method for flow based models.•We analyze the vulnerability of the proposed privacy defender.•We implement a real-time privacy protection system on embedded devices.