Fine-Grained Privacy-Preserving Image Retrieval in Cloud Environment
Jing Liang, Libo Wang, Peiya Li · 2024
With the advancement of cloud computing, an increasing number of resource-constrained data owners prefer to store their images in the cloud. Considering security and privacy, images should be encrypted before upload. Current privacy-preserving image retrieval techniques still face challenges such as high access control cost, insufficient granularity and low retrieval efficiency. Therefore, in this article we design an efficient fine-grained privacy-preserving image retrieval scheme based on polynomial access policies. Our scheme first employs transfer learning techniques to extract image features using pretrained deep learning models to enhance retrieval accuracy. Then an encrypted hierarchical K-means structure is established to cluster these features that improves retrieval efficiency without compromising data privacy. For fine-grained access control, a role-based polynomial access control technique is designed to enable precise permission management. Our scheme combines access control policies with index trees to greatly improve access speed in ciphertext domains. The practicality of the proposed scheme is demonstrated through experimental evaluations of its security and performance.