2026· IEEE Transactions on Network Science and Engineering· Vol 13, pp. 10496-10514· 1 citation· 105 references
Abstract
Achieving reliable network-wide consensus formation in distributed learning systems becomes increasingly challenging when edge nodes hold skewed data distributions. Federated learning (FL) enables privacy-preserving collaborative model training without sharing raw data, but statistical heterogeneity across nodes significantly degrades convergence and may expose sensitive label statistics to distribution-inference attacks. To address these limitations, this paper presents a new federated consensus-oriented aggregation (FedCOA) strategy, which improves consensus formation of FL under strong heterogeneity, suppresses skew-induced instability while mitigating distribution-level privacy leakage. FedCOA perturbs local label distributions using differential privacy (DP) and computes a noise-robust Index of Data Heterogeneity (IDH), which guides dynamic aggregation, regulates bias propagation, suppresses the influence of skewed updates, and facilitates consensus formation dynamics. We show theoretically that FedCOA reduces the divergence term in the convergence upper bound. Experiments demonstrate up to 80.3%, 75.1%, and 79.2% reductions in communication rounds on MNIST, FashionMNIST, and CIFAR-10, respectively.
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.
A. M., Nitish Kumar· International Journal of Mat...· 0 citations
This study addresses the challenges of privacy leakage and data silos in multi-source heterogeneous data interaction within smart grids by designing a federated multi-source data fusion architecture that combines adaptive local differential privacy with feature space alignment. This architecture utilizes Hessian matrix trace perception to adaptively adjust the local noise budget and introduces a dynamic aggregation selection mechanism based on the maximum mean difference, reconciling the conflict between differential privacy perturbations and feature manifold losses. Experimental results show that, while ensuring strict differential privacy boundaries, the system improves test accuracy by 7.45%, achieves a model inference speed of 45 FPS, and reduces communication resource overhead by 36.5%. Even under extreme conditions such as nonindependent identically distributed skew and 15% Byzantine poisoning attacks, it maintains a 98.40% attack interception rate and robust generalization fusion performance, providing a feasible system solution for building a highly reliable and resilient situational awareness and control foundation for the distribution IoT.
Jiaying Li, Can Pei· International Conference on...· 0 citations
Federated Learning (FL) avoids centralizing raw data, but server-side access to per-client updates still creates a significant privacy risk because gradients can leak sensitive information through inversion and related attacks. A common defense is client-level Differential Privacy (DP), which reduces attack fidelity by perturbing individual updates, but often at the cost of degraded predictive utility, especially in heterogeneous settings. In this paper, we study a different design choice: aggregate-only observability, where the server learns only an aggregate update rather than the individual client contributions that compose it. We develop a unified evaluation framework comparing baseline FL, client-level DP-FL, an idealized sum-only aggregation mechanism (hide_sum), and a quantum-inspired aggregate-observation primitive (qsum) based on finite-shot GHZ-style phase estimation. Experiments on Texas PUDF and MIMIC-IV show that the aggregate-only regimes preserve competitive predictive utility relative to baseline FL while removing the direct per-client observation channel used by standard server-side inversion attacks. We also analyze utility sensitivity to shot budgets and the resource scaling of the quantum-inspired primitive. The results suggest that privacy–utility tradeoffs in FL depend not only on how updates are perturbed, but also on what the server is allowed to observe. Aggregate-only observability therefore provides a useful mechanism-level complement to noise-based privacy defenses, while also motivating hybrid designs to address residual leakage from aggregate updates.
Clifford N. Jones, Md Nahid Hasan, S. Wagle et al.· International Conference on...· 0 citations
Federated low-rank adaptation (FedLoRA) allows multiple clients to collaboratively fine-tune large language models (LLMs) on downstream tasks without exposing their private data. To mitigate privacy leakage during aggregation, differential privacy (DP) is widely used to clip and perturb local model updates with noise, yet it can compromise model accuracy due to the inherent privacy-utility trade-off. The performance degradation becomes worse under the FedLoRA setting with the amplified DP noise impact and client heterogeneity in both model structure and data distribution. In this work, we propose iP-FedLoRA, a privacy-preserving federated fine-tuning framework for heterogeneous clients that strikes a good privacy-utility balance. Specifically, to fully utilize clients' heterogeneous resources, we customize LoRA modules based on their available resources. iP-FedLoRA employs matrix-wise differentially private local fine-tuning with sensitivity-aware noise allocation and rank-compensated LoRA regularization, which effectively alleviates noise impact of low-rank modules and enhances training efficiency. By leveraging noise-resilient knowledge distillation, iP-FedLoRA facilitates heterogeneous LoRA aggregation that selectively prioritizes high-confidence knowledge to filter DP-induced noise, thereby achieving robust knowledge transfer. Through rigorous privacy analysis and extensive experiments, we show that iP-FedLoRA provides privacy guarantees, improves model accuracy by up to 3.8%, and expedites training by 1.37-2.23×.
Nan Yan, Yu-Qing Li, Xiong Wang et al.· Proceedings of the 32nd ACM...· 0 citations
The fast growth of big data, along with the growing stricter privacy laws, poses tremendous challenges to the traditional centralized implementation of AI models. Federated Learning (FL) offers a decentralized solution; however, it is undermined by a number of performance limitations such as large communicational overhead, statistical heterogeneity due to nonindependent and identically distributed data, and small computational capability of edge clients. This paper presents an optimized AI-based FL architecture, which consolidates a client selection algorithm (that is based on reinforcement-learning) and a dynamic and fairness-conscious aggregation protocol, as well as a new query-fragment caching solution. The combined model is strategically focusing on high utility players, reducing skew in data during model aggregation, and lessening on-device unnecessary processing. The empirical analysis of CIFAR-10 and FEMNIST shows that the presented framework reduces the communication rounds by 38% and the model accuracy is increased by 5.7% when compared with Fed Avg and Fed Prox, at the same time, data privacy is also rigorously maintained.
Shashi Purna Kurapathi· International Conference Com...· 0 citations
In dynamic mobile decentralized federated learning (DFL), adversaries can poison both model updates and the topology information devices use to choose collaborators. We present DMTT (Dynamic MURMURA with Trusted Topology), a decentralized personalized FL protocol built on MURMURA, which uses evidential deep learning to down-weight distribution-mismatched peers, extended here to time-varying graphs under topology-manipulation attacks. Each device maintains a confidence-weighted local topology view from link-reliability estimates, signed topology claims, witness corroboration, and a Beta-distributed source-trust model, then aggregates only over a trust-screened collaborator set using a composite score fusing model compatibility, topology trust, and link reliability. We prove the screened mixing matrices confine Byzantine influence to a bounded residual $\delta_{max}$ that vanishes under perfect screening, and implement DMTT as a coordinator-free distributed system with each client running as an independent ZeroMQ process synchronized by a shared wall-clock epoch. On UCI HAR and PAMAP2, each partitioned across 100 mobile clients with Dirichlet heterogeneity, DMTT sustains honest-node accuracy above 0.862 (UCI HAR) and 0.829 (PAMAP2) across all tested adversary fractions (10 to 80%), nearly matching no-attack accuracy at low fractions and degrading gracefully toward local-only performance at extremes; static and dynamic FedAvg collapse to chance at every fraction, and robust aggregators (Krum, BALANCE, UBAR) fail to consistently beat a local-only baseline, while DMTT is the only method that clears this bar across both datasets at all fractions, with surviving Byzantine aggregation weight empirically zero throughout, consistent with $\delta_{max}$=0. The protocol runs end-to-end on real nodes via a coordinator-free ZeroMQ backend on the Melbourne Research Cloud.
Shubham Vaishnav, Murtaza Rangwala, Ali Beikmohammadi et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
