Privacy Protection for Image Sensors Using Convolutional Neural Network and Generative Adversarial Nets

An image privacy information protection method that does not damage the original image quality is proposed considering the risk of privacy information leakage due to the widespread use of image sensors and image sharing in social platforms. A convolutional neural network is used to detect the privacy information of the image. A generative adversarial network is then employed for the detected privacy content to protect image privacy by transforming the privacy information into non-privacy information without damaging the original image quality. Numerous studies have been conducted on real-world images., and results prove the efficiency and effectiveness of the proposed method.