Large-Scale Cross-Modal Ranked Search in Secure Intelligent Computing

With the development of intelligent industrial computing, cross-modal search technology has become crucial in fields, such as the Internet of things and industrial networks. Existing solutions, while supporting cross-modal retrieval and ensuring data privacy, suffer from high training complexity, low search efficiency, and vulnerability to attacks. To tackle these issues, this article presents a large-scale cross-modal search scheme designed for fast, efficient, and secure ranked retrieval. Specifically, we use collective matrix factorization to train heterogeneous data features and obtain unified feature vectors, simplifying the training process. Second, we design a label classification algorithm that employs homomorphic encryption and locality sensitive hash to improve search efficiency. In addition, we improve a secure method utilizing the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$k$</tex-math></inline-formula>-nearest neighbor algorithm to compute Euclidean distances, which effectively identifies results closest to the query data and resists linear analysis attacks. Both theoretical analysis and experimental results show that our proposed scheme not only protects the privacy of cross-modal data, but also offers high efficiency and practicality.