LAPA-based Dynamic Privacy Optimization for Wireless Federated Learning in Heterogeneous Environments

TL;DR

This work tackles privacy-preserving wireless federated learning under Non-IID data by introducing Lightweight Adaptive Privacy Allocation (LAPA), which personalizes DP budgets based solely on gradient information and training progress, avoiding extra communication. It couples LAPA with a dynamic noise control mechanism that shifts from artificial DP noise to environmental channel noise at a data-driven switching time $T_{th}$, with transmission power optimized by Deep Deterministic Policy Gradient (DDPG) to balance privacy and convergence. A Wasserstein-distance-based aggregation scheme, along with SINR-driven device selection, accounts for data and communication heterogeneity to improve aggregation quality. Theoretical convergence bounds are derived, and extensive simulations on MNIST and Fashion-MNIST demonstrate that LAPA substantially improves convergence while maintaining DP, outperforming FedAvg and prior DP strategies, and the dynamic noise optimization further enhances performance under realistic wireless conditions.

Abstract

Federated Learning (FL) is a distributed machine learning paradigm based on protecting data privacy of devices, which however, can still be broken by gradient leakage attack via parameter inversion techniques. Differential privacy (DP) technology reduces the risk of private data leakage by adding artificial noise to the gradients, but detrimental to the FL utility at the same time, especially in the scenario where the data is Non-Independent Identically Distributed (Non-IID). Based on the impact of heterogeneous data on aggregation performance, this paper proposes a Lightweight Adaptive Privacy Allocation (LAPA) strategy, which assigns personalized privacy budgets to devices in each aggregation round without transmitting any additional information beyond gradients, ensuring both privacy protection and aggregation efficiency. Furthermore, the Deep Deterministic Policy Gradient (DDPG) algorithm is employed to optimize the transmission power, in order to determine the optimal timing at which the adaptively attenuated artificial noise aligns with the communication noise, enabling an effective balance between DP and system utility. Finally, a reliable aggregation strategy is designed by integrating communication quality and data distribution characteristics, which improves aggregation performance while preserving privacy. Experimental results demonstrate that the personalized noise allocation and dynamic optimization strategy based on LAPA proposed in this paper enhances convergence performance while satisfying the privacy requirements of FL.

Figures (7)

• Figure 1: The workflow of proposed LAPA-based FL, involving only gradient exchange between devices and the BS, without incurring additional communication overhead, making it simpler and more efficient than the algorithm in hu2023shield.
• Figure 2: dynamic noise control mechanism.
• Figure 3: Device distribution in the simulation environment.
• Figure 4: Comparison of different aggregation strategies in a heterogeneous wireless communication environment.
• Figure 5: FL performance of different privacy allocation strategies under various artificial noise levels.