Aggregate-Only Observability for Privacy–Utility Tradeoffs in Federated Learning

URP-FL: Robust and Personalized Federated Learning under Heterogeneous and Adversarial Conditions

Federated learning is appealing for privacy-sensitive network systems, yet its practical deployment remains hindered by the following three recurring challenges: (1) client drift under non-IID data, (2) vulnerability to corrupted updates, and (3) the communication cost of repeated model exchange. Most existing approaches address these issues in isolation. While analytically convenient, this separation often fails to reflect real-world conditions. For instance, defenses against poisoning may suppress useful updates, while personalization and compression can alter the aggregation geometry itself. In this paper, we study these effects jointly and propose URP-FL, a compact training framework that integrates reliability-aware aggregation, local regularization for drift control, and sparse client uploads. We provide theoretical analysis establishing a convergence bound with distinct terms capturing optimization error, data heterogeneity, and adversarial impact. Experiments on a non-IID image classification benchmark with sign-flip and label-flip attacks demonstrate the benefits of the unified design. Compared to FedAvg and FedProx, this URP-FL maintains accuracy under attack while reducing transmitted parameters by approximately 75%. Rather than presenting a production ready system, it offers a reproducible and technically coherent step toward federated learning that is more robust under realistic conditions.

SNR-Guided Model Sizing for Differentially Private Federated Learning with LiRA Privacy Auditing

Federated Learning (FL) enables distributed training while keeping data local, but exchanged model updates can leak information through membership inference attacks. Differential privacy mitigates this risk via noise injection; however, aggressive DP regimes with strong noise can destabilize large models. An SNR-guided framework is introduced to select model dimensionality based on the signal-to-noise ratio imposed by the privacy budget. Three optimizers, DP-FedAvg, DP-FedAvgM, and DP-FedAdam, are evaluated across six domains, including image, clinical, IoT, and network security tasks. Privacy leakage is assessed using both loss-based membership inference and the likelihood-ratio attack LiRA. DP-FedAvgM achieves 98.10% accuracy on MNIST at ε =200 with LiRA AUC near random guessing (0.491). SNR-guided models reduce communication cost by up to 66×. Sensitivity calibration experiments further show that incorrect noise allocation can reduce accuracy by up to 1.36 percentage points. These results highlight the importance of model sizing and noise calibration for reliable privacy-preserving FL under strong DP constraints.

Federated Learning with Differential Privacy: A Comprehensive Framework for Privacy-Preserving Distributed Machine Learning

This study implemented a comprehensive experimental framework for analysing FL performance using standard FL aggregation protocols FedAvg, FedProx, and SCAFFOLD in conjunction with Differential Privacy mechanisms; specifically, the Gaussian noise mechanism with Rényi Differential Privacy (RDP) accountants.

Learning in the Right Subspace: Personalized Differential Private Federated Learning with Noise Filtering

Differential privacy (DP) mechanisms have been widely adopted in federated learning (FL) to enhance model security. However, existing approaches predominantly employ uniform privacy budgets, neglecting personalized requirements arising from heterogeneous user privacy preferences. Such uniform privacy configurations typically necessitate compliance with the most stringent budget, which not only leads to the wasteful underutilization of privacy budgets for certain clients but also compromises overall model utility. To address this limitation, we propose FedSPA, a Subspace Projection Aggregation personalized differential private Federated learning framework. The proposed method conducts singular value decomposition operations on noise-perturbed local models to extract singular value vectors as compact representations of both model structure and privacy noise. The server then clusters clients and identifies a consensus subspace for projecting models with varying noise levels, ultimately aggregating the global model through a residual-aware mechanism. This method not only effectively guides the aggregation of client personalized differential privacy but also reduces communication overhead. Extensive experiments demonstrate the model's effectiveness. Additionally, we provide theoretical proof of the privacy and convergence of FedSPA. Experimental results also showcase its superior performance over personalized DP-FL baselines.

Convex Optimization Methods for Privacy-Preserving Federated Learning

Federated Learning (FL) has emerged as a revolutionary paradigm in distributed machine learning, enabling multiple decentralized clients to collaboratively train models without sharing their local raw data. Despite its inherent privacy-centric design, FL remains vulnerable to sophisticated privacy attacks, such as gradient leakage and membership inference, which can reconstruct sensitive user data from communicated model updates. In order to reduce these vulnerabilities, we integrate privacy-preserving mechanisms most notably Differential Privacy (DP) and Cryptographic Protocols into the training procedure. These privacy constraints, however, come with utility loss and convergence slowdown thus highlighting a basic conflict between (differential) privacy on one side and high-order model accuracy and efficiency at another. In our paper, we carefully examine how to use convex optimization methods systematically in terms of performing this rich multi-dimensional trade-off. We center around the rigorous implementation of privacy-preserving FL couched as a bounded convex optimization task, studying how traditional and state-of-the-art optimization algorithms retain strong convergence rates even under durable privacy constraints. We benchmark the performance of these primary optimization frameworks, such as FedAvg, FedProx, and Accelerated Gradient Methods, when adopted on different privacy budgets. Theoretically, we analyze the impact of differential privacy on gradient variance in algorithms and experimentally validate how adaptive optimization (Specifically by AMSGrad) and proximal regularization can account for this noise-induced increase to enable faster convergence with a tight guarantee of differential privacy. To summarize, this work provides a unified approach for aiding the design of state-of-the-art privacy-preserving distributed learning systems that are also utility-optimal and is an important step towards using such approaches in high-stakes domains like healthcare or finance.

PrivSynth: Alternating and Control-Based Optimization for Privacy and Utility in Synthetic Data