The traditional federated learning (FL) approach in the Wireless Internet of Things (WIoT) relies on a central server to coordinate and manage the entire training process, which contradicts the design concept of distributed and autonomous IoT systems. In actual WIoT scenarios, it is difficult to find a reliable third party to serve as a central server to reduce the risk of single points of failure and system paralysis. Moreover, the dynamism and heterogeneity of the WIoT system require strong privacy protection mechanisms to safeguard the vast amounts of user data stored on devices. To tackle these issues, this paper introduces a decentralized federated learning (DFL) utilizing differential privacy (DP) applied in WIoT scenarios to ensure robust privacy protection. Furthermore, to address the black box model problem of FL neural networks, a one-point residual feedback (OPRF) mechanism is adopted. Unlike one-point feedback, regret analysis in OPRF relies on weaker assumptions and is more applicable to practical non-stationary optimization problems. Although OPRF increases randomness, complicating convergence boundaries and potentially causing error accumulation, rigorous analysis demonstrates that the proposed algorithm achieves a sublinear rate and$\epsilon$-differential privacy. Finally, simulation experiments validate the effectiveness and superior performance of the algorithm.

The Smart Healthcare Service Systems (SHSS) aim to integrate decentralized healthcare institutions, intelligent technologies, and end users into a cyber-physical system that enables high-quality medical decision-making. However, the sensitive nature of healthcare data presents significant privacy and security challenges, which hinder effective collaboration among healthcare providers. Moreover, existing research lacks a comprehensive theoretical framework that spans the full pipeline from data acquisition to intelligent decision services. To address these challenges, we propose a theoretical framework for SHSS that systematically analyzes data processing and user demand to establish the goals of secure, stable, and adaptive collaborative learning. Guided by these goals, a Decentralized Online Federated Learning (DOFL) network model is tailored for SHSS, where participating institutions interact through a decentralized federated learning structure. Building on this model, we design DP-DOOR (Differentially Private Decentralized Online Federated Learning with One-Point Residual Feedback), a fully decentralized algorithm that supports row-stochastic communication topologies, accommodating practical limitations where bidirectional synchronization is often infeasible. DP-DOOR ensures data privacy through differential privacy (DP) mechanisms and achieves efficient gradient estimation using a one-point residual feedback (OPRF) approach. Theoretical analysis shows that DP-DOOR provides ϵ-DP guarantees and achieves sub-linear regret. Experimental evaluations on diverse real-world medical datasets under both IID and non-IID settings demonstrate the algorithm’s robustness and effectiveness in enabling secure, decentralized collaboration and enhancing adaptability in dynamic healthcare environments.

In Healthcare Internet of Things Networks (HIoTNs), safeguarding the sensitivity of patients’ electric healthcare records (EHRs) is imperative, necessitating effective privacy protection while ensuring ample data for training. Decentralized federated learning (DFL), as a peer-to-peer distributed machine learning architecture, serves a pivotal role in preserving the data privacy of EHRs. However, two primary challenges in applying DFL to HIoTNs include safeguarding the privacy of EHRs at individual hospitals and optimizing communication resources among hospitals. This article proposes a three-layer healthcare framework comprising a physical layer, a communication layer and an edge-encrypted layer to segregate hospital local data from interaction models. Moreover, a novel doubly accelerated DFL algorithm (AccDFL) is introduced to integrate DFL training and information interaction into the communication layer of HIoTNs. By employing the heavy-ball and Nesterov methods, AccDFL achieves double acceleration, ensuring both$\epsilon _{i}$-differential privacy (DP) and linear convergence. The theoretical analysis of AccDFL delves into the interaction among distributed gradient tracking (DGT), DP, and accelerated mechanisms, presenting comprehensive convergence and privacy analyses to overcome the exponential increase in parameters brought by the acceleration and DP mechanisms. Experimental results with realistic non-IID grayscale and color medical datasets of different disease types affirm the significant advantages of AccDFL over other algorithms in terms of accuracy and communication efficiency.

Heart disease is a leading global cause of death, while federated learning (FL) is an effective way to predict it. Due to patient privacy concerns and the centralized nature of current FL approaches, collaborative research in heart disease prediction (HDP) faces significant hurdles in computational health systems. This article introduces a distributed online aggregation method within fully decentralized federated learning (DFL), named DeFedHDP, to address these privacy challenges and improve the model of HDP. Moreover, the differential privacy (DP) mechanism is applied to the aggregation strategy of DeFedHDP to protect the privacy of the patients. The data holder communicates directly with neighbors in a series of time-varying directed graphs without the involvement of a central server. Furthermore, each participant is both a trainer of the local model and a collaborator of the other participants’ models. The data does not leave the local device, only the model parameters are exchanged and integrated, and this decentralized approach can further improve the level of privacy protection. In addition, to cope with model gradient disappearance and gradient explosion, the one-point bandit feedback (OPBF) strategy is utilized to estimate the true gradient values. Experiments on a public medical dataset show that the effectiveness of DeFedHDP is close to the centralized FedAVG algorithm for client-server architectures in terms of accuracy and speed.