Multi-user privacy-preserving image search in distributed cloud computing

作     者：Zhang, Zongye Zhou, Fucai Wang, Yan 

作者机构：Northeastern Univ Software Coll Shenyang 100819 Peoples R China 

出 版 物：《COMPUTERS & ELECTRICAL ENGINEERING》 (计算机与电工)

年 卷 期：2024年第118卷第PartB期

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：National Natural Science Foundation of China [62072090, 62202090, 62173101] Liaoning Province Natural Science Foundation Medical-Engineering Cross Joint Fund [2022-YGJC-24] Doctoral Scientific Research Foundation of Liaoning Province [2022-BS-077] Fundamental Research Funds for the Central Universities [N2417006] 

主　　题：Encrypted image search Privacy-preserving Searchable encryption Distributed cloud computing 

摘      要：Cloud-based image search has emerged as a research focus, but security and privacy concerns limit its application. One secure solution is to encrypt the images before outsourcing them to the cloud. Nevertheless, encryption significantly reduces image availability and suffers from some privacy leakages, such as search pattern leakage, access pattern leakage, and similarity leakage. With the exponential growth of images, the distributed cloud architecture is adopted to provide storage and search services for vast amounts of images. However, ensuring multi-user access control in distributed cloud computing remains a challenge. To address these problems, we propose a multi-user privacy-preserving image search scheme in distributed cloud computing environments. First, we propose an SGX-based privacy-preserving distributed image search framework that leverages SGX to protect image similarity against cloud servers. Second, we present an index re-encryption strategy and an access obfuscation strategy to preserve the search pattern and the access pattern. Last, we propose a multi-secret sharing-empowered multi-user access control method enabling multi-user queries and facilitating lightweight multiuser management in distributed environments. Security analysis and experimental evaluation demonstrate that our proposed scheme guarantees privacy with efficient search.