Differential Privacy-Preserving of Multi-Party Collaboration Under Federated Learning in Data Center Networks

The unit of federated learning and differential privacy protection technology can ensure that the published aggregation has sufficient anonymity. While realizing privacy protection, users' data can be fully utilized. However, attackers may still steal user information by eavesdropping on the shared model of federated learning participants. In this paper, we mainly study the multi-party collaborative privacy protection mechanism based on federated learning. Firstly, we design a stochastic forest algorithm satisfying differential privacy in a distributed computing framework. The algorithm first segments the nodes in the weak classifier into the best segmentation points, and then perturbs the leaf nodes through the Laplace mechanism to protect the classification results. Secondly, we present a differential privacy-preserving mechanism based on XGBoost, aiming at the privacy leakage problem that may occur during the construction of weak classifiers. Finally, we propose a multi-party collaborative construction scheme based on federated learning to meet differential privacy. At last, we empirically confirm that our algorithm achieves state-of-the-art (SOTA) performance compared with existing SOTA baselines in DPFL. Moreover, numerous experimental results indicate that the proposed framework can ensure data security and also have high accuracy.