CAPIA: Cloud assisted privacy-preserving image annotation

Using public cloud for image storage has become a prevalent trend with the rapidly increasing number of pictures generated by various devices. For example, today's most smartphones and tablets synchronize photo albums with cloud storage platforms. However, as many images contain sensitive information, such as personal identities and financial data, it is concerning to upload images to cloud storage. To eliminate such privacy concerns in cloud storage while keeping decent data management and search features, a spectrum of keywords-based searchable encryption (SE) schemes have been proposed in the past decade. Unfortunately, there is a fundamental gap remains open for their support of images, i.e., appropriate keywords need to be extracted for images before applying SE schemes to them. On one hand, it is obviously impractical for smartphone users to manually annotate their images. On the other hand, although cloud storage services now offer image annotation services, they rely on access to users' unencrypted images. To fulfill this gap and open the first path from SE schemes to images, this paper proposes a cloud assisted privacy-preserving automatic image annotation scheme, namely CAPIA. CAPIA enables cloud storage users to automatically assign keywords to their images by leveraging the power of cloud computing. Meanwhile, CAPIA prevents the cloud from learning the content of images and their keywords. Thorough analysis is carried out to demonstrate the security of CAPIA. A prototype implementation over the well-known IAPR TC-12 dataset further validates the efficiency and accuracy of CAPIA.