Edge Computation Network Resource Scheduling and Privacy Protection Mechanism Based on Federated Learning

Abstract

The diversity of edge devices and unpredictable network environments, along with the quick expansion of Internet of Things devices, have resulted in ineffective resource scheduling and privacy leakage threats. To address resource contention issues in dynamic environments, ensure security protection on resource-constrained edge devices, and improve the efficiency of resource scheduling and privacy protection in edge computing networks, this study designs a network computing resource optimization and data security protection solution based on federated learning. The research method deeply integrates federated learning with edge computing, adopts hierarchical federated learning technology to solve resource scheduling problems, and uses the Lagrange optimization method to achieve a closed-form solution for resource scheduling. By combining local differential privacy and homomorphic encryption technologies, a noise injection, masking mechanism, and ciphertext computation scheme are designed to ensure data privacy and security. The outcomes indicated that the task completion delay of the research method was 1250 ms when the data volume was 100 GB. Furthermore, when the privacy budget increased from 0.1 to 10, the efficiency value decreased by only 0.26. In practical application testing, the communication volume of the research method was 1.5MB per round when the number of clients was 100. Additionally, when the network fluctuation level increased from level 1 to level 5, the dropping rate increased by only 3.3%. The above results indicate that the resource scheduling and privacy protection mechanism based on federated learning in edge computing networks, as proposed by the research, is highly practical, robust and secure. It effectively solves the problems of low resource scheduling efficiency, poor adaptability and inadequate privacy protection capabilities in edge computing networks.