Overview of Privacy Set Intersection Protocol Based on Heterogeneous Network and Social Network User Alignment

First, embed the knowledge graph into the user alignment method, vectorize the data information features, and project them onto the same feature space. Then, combining heterogeneous networks and social network environments, used data information and network structure to establish a similarity matrix for users, and used the similarity matrix to determine whether they point to the same user, optimized and derived the objective function based on matrix theory, added regularization terms to avoid too much difference between the alignment matrix and the prior alignment matrix, and solving it. Once again, based on the user alignment concept mentioned above, a privacy intersection protocol based on fully homomorphic encryption and hash algorithm was proposed in a privacy environment, which used fully homomorphic encryption and hash algorithm to encrypt data to ensure security. Finally, in order to design a multi-party privacy intersection protocol that adapts to multi-party interactions, Freedman based two-party/multi-party intersection protocols had been proposed, and the multi-party secure intersection protocol was tested, with an average accuracy of 90.3%. User alignment and privacy intersection are prerequisites for personalized computing, and also provide a theoretical basis for the development of Vertical Federated Learning and privacy computing.