A privacy-preserving image retrieval scheme based secure kNN, DNA coding and deep hashing

Abstract

In recent years, information leakage incidents occur frequently. Ciphertext domain search has become a hot spot in multimedia technology and software engineering. Therefore, how to mine valuable information safely and effectively is very important. In this paper, we present the CBIR solution and complete the relevant software development. The proposed method not only outsources CBIR services but also avoids privacy leaks. Our main contributions are reflected in two aspects: improving search efficiency and protecting user privacy. For one thing, we propose a novel privacy protection algorithm to encrypt images. For the other thing, we propose an integrated deep hash algorithm to extract the high-level features of images. In the index encryption process, we use secure kNN to encrypt the index. Through experimental analysis and argumentation, the proposed privacy protection algorithm has a lower correlation coefficient and a better information entropy. It is secure enough and can resist common types of attacks encountered during data transmission. The proposed secure image retrieval algorithm is tested on two datasets. Compared with the same type of algorithm, the proposed algorithm has better retrieval performance. In short, the privacy reservation scheme guarantees the security and effectiveness of the system.