Removing sensitive parts of an image

Abstract

Social media such as Facebook and Instagram has gathered more than 2 billion individual users from all around the world. With the convenience of uploading photos online, photo sharing is one of the growing form of communication in the 21st century. However, with more people sharing photos online, there is also a higher risk in oversharing or sharing sensitive information online. As such, online photo sharing raises concerns as it may unintentionally disclose sensitive information in an image. In this paper, we study the implications of negligently sharing photos with sensitive data on social media; as well as some of the research carried out and features proposed in recent years to protect image privacy. To provide a safe environment for responsible users when sharing photos online, we developed a tool which provides user a solution for image privacy protection. It is achieved by: 1) Performing image segmentation using convolutional neural network by pairing detected objects to a class tag 2) Allowing users to blur a specific sensitive object in the image 3) Allowing users to replace a specific sensitive object with another similar object in the image to protects the image’s confidentiality. We targeted 21 classes of common daily life objects which will be detectable and can be replaceable and/ or blurred depending on a user’s preference.