Research on Multi-source Data Security Fusion Analysis Technology Based on Federated Learning

In this paper, we study the multi-source data security fusion analysis technology based on federated learning. A rapid model convergence technique based on data distribution difference evolution comparison is proposed to explore the data distribution law among participants and accelerate the model convergence speed. In view of the difference of data distribution among participants, the difference between local data distribution and global data distribution is obtained by analyzing the evolution law of data distribution of participants. Then, the weight is assigned according to the similarity between the data distribution of each participant, and it is used to guide the rapid convergence of the global global model, and improve the convergence speed of the machine learning model based on the federated learning framework. A collaborative optimization technology based on cooperative game is proposed, which takes into account the performance of individual modeling and global modeling, and improves the accuracy of the model. The accuracy of the global model and the accuracy of each participant's local personalized model are considered for scenarios that need to meet both the participant's personalized modeling and global modeling requirements. Through the cooperative game model, the accuracy optimization of the global model and the local personalized model of each participant is solved, and the coordination of the accuracy of the global model and the local personalized model is realized.