A Searchable Encryption Technique for Secure Color Image Retrieval

Image retrieval is one of the fundamental tasks in computer vision that involves searching and retrieving relevant images to a query from a large collection of images. It has applications in various domains, including online education, image search engines and disease diagnosis. For image owners, it is often cost-effective to avail third-party owned computation and storage resources such as a cloud paradigm, to facilitate their image retrieval application. However, in addition to the associated privacy risks during transmission, storage and computation; image owners lose control over their data, which is not desirable, especially for privacy sensitive data. Therefore, this work proposes a searchable encryption scheme that protects images in a way that their features can be extracted in the encryption domain to ensure privacy. Also, it allows image owners to encrypt images with their own keys before uploading them to the cloud-server, thus leaving them in control of their data. In addition, we propose a feature fusion technique that combines fuzzy logic and color micro-structures to define an image in a high-dimensional feature space for retrieval application. The simulation analysis shows that for non-secure image retrieval, proposed fusion technique delivered up to 3% better precision score. On the other hand, our proposed searchable encryption method degraded precision up to 2% at best for secure image retrieval applications compared to its non-secure counterpart.