GAN-Based Image Privacy Preservation: Balancing Privacy and Utility

Zhenfei Chen, Tianqing Zhu, Chenguang Wang, Wei Ren, Ping Xiong

Abstract: While we enjoy high-quality social network services, image privacy is also under great privacy threats. With the development of deep learning-driven face recognition technologies, traditional protection methods are facing challenges. How to balancing the privacy and the utility of images has been an urgent problem to solve. In this paper, we propose a novel image privacy preservation model using Generative Adversarial Networks(GANs). By generating an fake image that highly matches the key face attributes of the original image, we balance the privacy protection and utility.