Privacy Challenges and Solutions for Image Data Sharing

Sharing image data benefits a wide range of applications, including social media, medical imaging, and intelligent systems. Image data often contain sensitive information, the sharing of which may inflict individual privacy concerns. Traditional image privacy techniques, such as pixelization and blurring, do not provide effective protection. In this paper, we discuss privacy challenges and solutions for image data sharing. Specifically, we review existing solutions based on cryptography and federated learning, and discuss recent results on differential privacy in image domain. While differential privacy provides provable guarantees, we identify specific privacy challenges for image data and point out several considerations for future research.