A heterogeneous network structure publishing security framework based on cloud-edge collaboration

With the integration of different fields, a large amount of heterogeneous data converges into a heterogeneous network with different types of nodes and edges. Then it is published by the cloud platform for applications such as recommendation services and public opinion analysis. However, in traditional cloud platform data publishing, one is privacy leakage due to the long transmission delay to the cloud platform, and the other is privacy leakage due to privacy attacks during cloud platform data publishing. In addition, there is an imbalance between privacy and availability in the privacy protection of the network structure between heterogeneous nodes. To address these issues, we propose a heterogeneous network structure publishing security framework based on cloud-edge collaboration. In this framework, we design a non-interactive edge privacy protection center and provide a heterogeneous network data publishing privacy protection model (HNPP). First, the network between heterogeneous nodes is transformed into an equivalent homogeneous network based on trivial closure. Then, we use differential privacy and random perturbations to achieve the homogeneous network structure reconstruction. Finally, we only rely on generation rules and reconstructed equivalent homogeneous networks to generate the heterogeneous network structure. We conducted extensive experimental analysis on real datasets, and the results show that the original edge retention rate of the network after privacy protection is above 0.88, and the clustering precision is above 0.92. It shows that the HNPP model is able to balance the privacy and availability of heterogeneous network structure publishing.