The importance and need for data privacy is now commonly discussed everywhere and many regulations are also in place to ensure privacy preservation. Personal information like age, gender, marital status, health records etc. are the common data which needs privacy. However, images also contain sensitive information that can compromise individual privacy. With huge number of images being uploaded into public domain, image privacy also became a major privacy concern. The conventional methods of image privacy include image obfuscation which results into images which are not useful enough for further analytics. Modern literature suggests usage of Generative Adversarial Networks (GAN) to generate synthetic data for image privacy protection. The tradeoff between privacy and utility still exists. Our solution is customized GAN based method which is able to generate synthetic face images which can strike a balance between privacy and utility. We have used structural similarity index measure (SSIM) to compare the images generated using various methods and our image privacy method achieved SSIM of 98.3% along with metadata being preserved. The following sections will describe our contributions in detail.

Computer vision, image privacy, Generative Adversarial Networks, Structural Similarity Index Measure, Privacy Preservation