Research on cross-border enterprise knowledge sharing mechanism based on federated graph neural network and differential privacy collaboration

With the acceleration of globalization, cross-border knowledge sharing among enterprises has become a key factor in enhancing international competitiveness. However, increasingly prominent data privacy and security issues have limited its development. The combination of Federated Graph Neural Networks (FGNN) and Differential Privacy (DP) technology offers a novel approach to resolving this contradiction. This study proposes a cross-border enterprise knowledge-sharing mechanism that collaborates with FGNN and DP to achieve efficient knowledge transfer while ensuring data privacy and security. By constructing a multi-node federated learning framework, a graph neural network is used to capture complex knowledge associations among enterprises, and differential privacy is employed to mask gradient information, effectively preventing the leakage of sensitive information. The experiment uses the cross-border enterprise cooperation dataset, and different privacy budgets (ε = 0.1, 0.5, 1.0) are employed to compare the model’s performance. The results show that when ε = 0.5, the model’s accuracy in the knowledge-sharing task reaches 87.3%, which is 12.6% higher than the baseline federated learning model and meets the privacy protection requirements. In addition, when the number of nodes increases to 50, the collaboration mechanism can still maintain an accuracy rate of 83.1%, which verifies its scalability. Compared to existing cutting-edge methods, this mechanism achieves a dual breakthrough in privacy protection and knowledge sharing efficiency, providing a new solution for cross-border data compliance collaboration.