Hybrid fractional groupers and moray eels driven deep learning for pneumonia detection using multi-modal data in federated learning.

ATSF-SFL: Adaptive spatiotemporal feature fusion and split federated learning framework based on rs-fMRI for multi-center brain disease diagnosis

The advancement of cutting-edge technologies such as the Internet of Things (IoT) and Deep Learning (DL) has transformed the Internet of Medical Things (IoMT) based healthcare into a new paradigm known as the Healthcare 5.0. This paradigm shift, particularly within Healthcare 5.0, introduces smart, cost-effective, and sustainable healthcare services. However, in such complex and heterogeneous IoMT based networks, smart devices generate large volumes of imbalance data. Most DL models struggle to accurately distinguish malicious behavior and, consequently, fail to detect network threats effectively. To address these challenges and ensure privacy preservation, we propose a novel Generative Adversarial Network (GAN)-oriented Federated Learning (FL) model. The proposed approach generates realistic synthetic data for improving the detection of minority-class threats, while FL facilitates distributed training without revealing raw data. Additionally, a Bidirectional Long Short-Term Memory (BiLSTM) network is employed to identify various attack types within smart IoMT-based Healthcare 5.0 systems. Experimental results on two benchmark imbalance datasets, UNSW-NB15 and NSL-KDD, demonstrate that the proposed model achieves superior accuracy of (94.78% and 95.90%) and F1-score (94.88% and 98.70%) respectively for minority-class attacks, outperforming existing methods.

Federated Learning (FL) in edge-enabled Internet of Things (IoT) networks faces considerable challenges owing to intermittent client participation and distributional drift, and which undermines the stability of a global model's optimization. This coupled impact introduces temporal unreliability, in turn, impairing the training stability. State-of-the-art FL frameworks typically address these challenges in isolation and overlook their coupled impact particularly during the client reintegration process. In order to address this limitation, we propose BRIDGE-T, i.e., a reliability-aware FL framework that addresses temporal unreliability in edge-enabled IoT networks. BRIDGE-T encompasses three components, i.e., (i) Prototype Contrastive Drift Alignment (PCDA) to constrain cross-client representation divergence under evolving non-Independent and Identically Distributed (non-IID) data, (ii) Prototype Query Agreement (PQA) to estimate round-wise clients reliability via cross-client prediction consistency on shared prototypes, and (iii) Reliability-Weighted Asynchronous-aware Aggregation (RWAA) to regulate clients' influence and attenuate stale or misaligned clients' updates. Extensive experiments under varying intermittency and distributional drift on CIFAR-10, CIFAR-100, MNIST, and TON-IoT suggest that BRIDGE-T achieves smoother convergence and greater robustness to client reintegration vis-à-vis the state-of-the-art FL frameworks.

The rapid growth of the Internet of Things (IoT) has introduced significant security vulnerabilities and increased the risk of cyberattacks. Intrusion Detection Systems (IDS) are widely used to identify malicious activities; however, their detection accuracy is often degraded by latency and privacy concerns in centralized environments. Federated Learning (FL) has therefore emerged as a privacy-preserving solution for distributed intrusion detection. The FL-based intrusion detection remains challenging due to data imbalance across distributed nodes and susceptibility to adversarial attacks, which can degrade model generalization and robustness. To address these challenges, this paper proposes a Groupers Brown Bear Optimization-based Spiking Residual ShuffleNet (GBOA_SR-ShuffleNet) framework. In the proposed approach, the GBOA algorithm is used to optimally train the SR-ShuffleNet, enabling improved parameter tuning under heterogeneous and imbalanced data distributions typical of FL environments. This leads to more stable model updates, convergence and enhances robustness against adversarial effects, thereby improving the reliability of federated intrusion detection. The servers and IoT nodes are the main entities of the FL-based intrusion detection framework. In local training, intrusion detection is carried out, where the data are normalized by Dual normalization to stabilize data distribution and improve learning convergence. The features are fused using the Kumar-John distance measure with Deep Kronecker Network (DKN), which enhances discriminative feature representation and reduces redundancy. The Bootstrapping method augments the data to avoid class imbalance, and intrusion detection is performed using SR-ShuffleNet. The GBOA trains the SR-ShuffleNet, and Shapley Additive xPlanations (SHAP) show the final result of intrusion detection, which is utilized to provide interpretability and explain the detection decisions. Moreover, the GBOA_SR-ShuffleNet attains the accuracy, Mean Average Precision (mAP), loss, Mean Squared Error (MSE), Root MSE (RMSE), Root Relative Squared Error (RRSE), recall, F1-Score, and False Alarm Rate (FAR) of 96.48%, 95.63% 0.035, 0.080, 0.282, 0.336, 96.93%, 96.28%, and 3.15%.

FedCAD: Cross-modal semantic alignment and distillation for cross-domain heterogeneous federated learning.

Named Data Networking (NDN) represents a paradigm shift toward content-centric architectures but remains critically vulnerable to Interest Flooding Attacks (IFAs), where malicious actors overwhelm router Pending Interest Tables with spurious requests, causing service degradation and denial-of-service. To address the limitations of existing approaches, including high false positives in threshold-based methods and substantial overhead in centralized learning, we propose FL-IFAshield, a novel federated learning framework for adaptive IFA mitigation. Our solution integrates dynamic Poisson-EMA thresholding for accurate flood detection, entropy-aware federated aggregation to handle non-IID traffic distributions across edge routers, and Byzantine-robust mechanisms with differential privacy guarantees. Comprehensive evaluation on the FIT/IoT-LAB testbed with 100 routers demonstrates exceptional performance: 93.1% F1-score in attack detection, only 5% false positives, 28 ms average end-to-end latency ([Formula: see text]), and over 90% legitimate Interest Satisfaction Ratio under sophisticated collusive attacks, while maintaining minimal computational overhead (<9% CPU utilization on ARMv8 routers). FL-IFAshield significantly improves security performance, offering 35% higher accuracy than static thresholding and 60% lower communication overhead than centralized approaches. While simpler heuristic baselines naturally incur marginally lower computational footprints, our solution delivers the optimal overall operational balance among high precision, low end-to-end latency ([Formula: see text]), and resource efficiency in constrained edge computing environments.

Federated learning (FL) has emerged as a promising paradigm for accelerating magnetic resonance (MR) image reconstruction while preserving data privacy in multicenter collaborations. However, existing FL-based reconstruction methods face two major challenges: 1) a heavy reliance on fully sampled k-space datasets for model training, which is often a tricky problem in clinical settings, and 2) significant performance degradation due to distribution shifts between training and test domains. To address these limitations, a test-time personalization self-supervised FL (TTP-SSFL) method is proposed to accelerate MR image reconstruction. In this study, cross-institutional collaboration without any fully sampled data is implemented by introducing a Siamese-based self-supervised strategy with a hybrid loss function at each client. Moreover, a low-rank adaptation (LoRA)-based test-time adaptation (TTA) strategy is proposed to further mitigate domain shift during deployment. By inserting lightweight adapters into the global model and optimizing them using only testing data via self-supervision, the proposed method can achieve efficient model personalization and robust generalization under distribution shifts. Extensive experiments on multicenter datasets show that TTP-SSFL achieves state-of-the-art performance among self-supervised methods and matches the accuracy of supervised personalized FL (PFL) models, providing a practical and privacy-preserving solution for robust MR reconstruction across heterogeneous clinical environments.

Diabetic Retinopathy (DR) is still a major cause of vision loss that can be avoided. This means that we need automated screening systems that can work across institutions without putting sensitive medical data in one place. Although Federated Learning (FL) allows for cooperative model training while preserving data locality, Non-IID data distribution, communication overhead, and unstable convergence frequently limit its effectiveness in medical imaging. This paper suggests a federated Mixture-of-Experts (FL-MoE) framework for DR classification that combines interpretable deep learning and expert specialization in order to overcome these challenges. Using the EyePACS and APTOS-2019 retinal fundus datasets, this paper evaluates multiple backbone architectures, including Convolutional Neural Networks (CNN), a hybrid CNN-LSTM model, and transformer-based Vision Transformer (ViT), within the FL-MoE framework. FL-MoE improves performance under heterogeneous client distributions for several backbone architectures, particularly CNN-LSTM, though performance varies across models. The CNN-LSTM backbone achieves 76.2% accuracy with 89.5% AUC on EyePACS while reducing communication cost by an order of magnitude compared to transformer-based models. Furthermore, CNN-LSTM exhibits more stable convergence and stronger robustness to client-level data heterogeneity. Grad-CAM based explainability analysis qualitatively shows attention maps highlighting retinal regions commonly associated with DR. To quantify localisation quality, we computed Intersection-over-Union (IoU) with IDRiD lesion masks; mean IoU values were below 0.03 for all lesion types, confirming the coarse, exploratory nature of the visualisations. Overall, the proposed FL-MoE framework with a CNN-LSTM backbone offers an effective and practical solution for scalable, privacy-aware Diabetic Retinopathy screening in federated clinical environments, outperforming both standard federated baselines and a representative personalized FL method (FedBN) under heterogeneous data conditions.

Federated Learning in Endodontics: A Framework for Privacy-Preserving Multicentre Artificial Intelligence.

Electroencephalography (EEG) records electrical brain activity from the scalp and is widely used in brain-computer interface (BCI) systems for communication, and assistive technologies. EEG is widely used in motor-imagery (MI) based BCIs, where neural recordings contain highly individual and potentially sensitive information. In this regard, federated learning (FL) is a prominent privacy-enhancing approach which enables collaborative model training without centralising raw signals. However, recent work has shown that FL models still leak private information through membership inference attacks (MIAs). Most existing studies examine only single attack type, so it remains unclear how multiple MIAs together expose different layers of privacy risk in FL-based EEG systems. To address this gap, this study develops a federated MI-EEG classification framework and evaluates privacy leakage across four complementary MIAs: record-level, feature-level, gradient-level, and client-identity inference. Two neural networks were trained using per-subject FL, and differential privacy (DP) with epsilon (ε) ∈ {1, 5, 10} was applied to client updates. Results showed that standard FL alone provides limited intrinsic protection, while adding DP substantially reduces attack success particularly for gradient and identity-level attacks. Strong privacy settings (ε = 1) offered the greatest leakage reduction but degraded classification accuracy, whereas a moderate privacy budget (ε = 5) achieved the most favourable privacy-utility balance. Overall, the findings demonstrate that FL alone is insufficient as a privacy safeguard for EEG-BCI systems. Explicit privacy mechanisms such as DP are required to mitigate multi-level leakage, supporting the design of trustworthy and secure neural-learning technologies.

Many organisations collect sensitive data that cannot be freely shared. Hospitals store brain magnetic resonance imaging (MRI) scans on internal servers; banks keep transaction records behind strict firewalls; agricultural services retain crop images in isolated repositories. Federated learning (FL) allows models to be trained without centralising raw data, yet most existing systems address a single domain and offer limited insight into model behaviour and provenance over time. BlockFedX is a cross-domain federated learning system designed to address three simultaneous tasks: credit card fraud detection on tabular data, brain tumour detection on MRI images, and plant disease recognition on leaf images. These three domains were deliberately selected because they represent the principal data modalities in real-world privacy-sensitive deployments-structured tabular records, greyscale medical images, and colour natural images-and because public benchmark datasets exist for all three, enabling reproducible evaluation. The system uses a shared backbone that is updated only where model layers have compatible tensor shapes, while domain-specific output layers remain local at each client. Explanations are computed at the clients using SHAP feature-attribution for tabular data and Grad-CAM visual heatmaps for images; the server receives only compact statistical summaries. The server also applies a distance-based anomaly test on client updates and records model hashes, explanation summaries, and anomaly flags in a hash-chained ledger. Experiments on three public datasets under non-identical client data distributions show that BlockFedX achieves an average fraud-detection F1-score of 0.92, 74.32% mean validation accuracy on BrainMRI, and 77% test accuracy on PlantVillage, while keeping all raw data local. These results are below strong centralised baselines, as expected under compact models and non-IID splits, but the system simultaneously provides three properties rarely combined in prior work: cross-domain federated training via a shape-safe backbone, client-side explanations integrated into the learning loop, and a lightweight tamper-evident record of model evolution across rounds.

Deep-learning-based human activity recognition (HAR) has been widely studied and applied in recent years, but it raises privacy concerns. Federated learning (FL) enables collaborative training without sharing raw data, thereby protecting user privacy. However, FL for HAR is challenged by three coupled factors: non-IID data across clients, aggregation under heterogeneous local models, and stringent computation and bandwidth budgets on edge hardware. To address these factors, this work introduces FedSynHAR, a lightweight FL framework that combines Gradient-Importance-based Adaptive Pruning (GIAP) with Channel-guided Feature-level Mutual Distillation (CFMD). GIAP prunes both server and client networks based on gradient importance, reducing computational and communication overhead; CFMD uses channel importance to guide mutual distillation between local and proxy models and mitigates pruning-induced degradation to improve robustness under non-IID conditions. Experiments on UCI-HAR and PAMAP2 demonstrate the effectiveness of FedSynHAR for federated HAR. On UCI-HAR, FedSynHAR converges about 2× faster than FedAvg, achieves 94.91% accuracy under non-IID settings, and reduces overhead by up to two orders of magnitude. Results on PAMAP2 further support the robustness of FedSynHAR under stronger heterogeneity.

Deep learning-based computer-aided diagnosis (DL-CAD) models have achieved remarkable success in X-ray image analysis. Yet their effectiveness is often constrained by the tight regulations of governing sensitive X-ray data. Federated Learning (FL) offers a promising paradigm by facilitating the collaborative training of DL-CAD models across healthcare institutions without compromising data privacy. Despite this advancement, the limited local X-ray archives and the issue of heterogeneous model architectures bring distinctive challenges. To address these challenges, this work pioneers the utilization of natural images to develop a natural image-augmented heterogeneous FL framework (NatIMG-FL) for X-ray classification. For augmenting local training, NatIMG-FL leverages natural images as auxiliary supervised data to facilitate the alignment of feature distributions between natural and X-ray images. To tackle the model heterogeneity issue, NatIMG-FL introduces a novel dual weights-based fine-grained knowledge transfer method, enabling adaptive knowledge exchange between local and central models. The NatIMG-FL framework provides insights into exploiting natural images as proxy datasets to enhance knowledge transfer in heterogeneous FL for X-ray classification.

Artificial Intelligence (AI) models are critical for detecting advanced Internet of Things (IoT) botnets. However, these systems are highly vulnerable to Adversarial Machine Learning (AML), where malicious inputs are crafted to cause misclassification (e.g., identifying malicious traffic as benign), posing a systemic threat to IoT security. This systematic literature review (SLR) addresses the persistent "reality gap" between theoretical AML research, often derived from computer vision, and the practical, domain-specific constraints of network security. This paper synthesizes research from 2020–2025, providing comprehensive taxonomies of: (RQ1) targeted AI models, from traditional ML to modern Federated Learning (FL) frameworks; (RQ2) attack methodologies, highlighting the shift from feature-space (e.g., PGD) to realistic problem-space attacks (e.g., binary diversification, XAI-based attacks); (RQ3) proactive (e.g., Adversarial Training) and reactive defense strategies; and (RQ4) evaluation frameworks, critiquing the use of outdated datasets. Finally, (RQ5) we analyze open challenges, focusing on the IoT resource-constraint dilemma—where effective defenses like Adversarial Training are too computationally expensive for edge devices —and performance trade-offs. We conclude by outlining future directions, emphasizing the need for constraint-aware defenses, secure FL, and leveraging Generative AI.

Traditional synchronous Federated Learning (FL) is subject to the waiting latency inherent to synchronization mechanisms. Consequently, its convergence rate is constrained by straggler nodes within heterogeneous environments. Asynchronous Federated Learning (AFL) improves execution efficiency by removing global synchronization barriers. However, when integrated with blockchain for decentralized deployment, it still encounters challenges such as on-chain storage overhead arising from model parameters, convergence perturbations induced by stale gradients, and Byzantine security threats. To this end, this paper proposes BCAFL, a decentralized blockchain framework tailored for semi-asynchronous federated learning. BCAFL utilizes the InterPlanetary File System (IPFS) to implement off-chain storage for global model parameters. By integrating Model-Agnostic Meta-Learning (MAML) and PowerSGD, the framework enhances the model's local adaptation capability on non-IID data while concurrently reducing communication overhead. To safeguard model security and convergence stability in asynchronous environments, this study develops a Mutual Information and Delay-Aware (MIDA) dynamic aggregation mechanism. This mechanism leverages Mutual Information (MI) to perform model verification for defense against poisoning attacks, while simultaneously modulating aggregation weights via a dynamic aggregation factor to effectively mitigate model oscillations inherent in asynchronous convergence. Additionally, this study develops a dynamic stake-based Verifiable Random Function (VRF) committee consensus mechanism. By quantifying election weights based on node contributions, this approach enhances consensus efficiency and resistance to Sybil attacks. Simulation results demonstrate that, compared with various existing baseline schemes, BCAFL maintains the convergence accuracy of the global model while reducing communication overhead. It effectively suppresses convergence oscillations caused by asynchronous delays and defends against poisoning and Byzantine attacks. Furthermore, when the network scale is expanded to 300 nodes, the consensus latency does not show a significant increase.

From algorithmic innovation to clinical deployment: A systematic review of methodological gaps limiting federated learning in healthcare.

FedLASE: Performance-balanced system-heterogeneous FL via layer-adaptive submodel extraction.

Industrial Internet of Things (IIoT) ecosystems are expanding rapidly. Scalable and reliable intrusion detection systems (IDS) are needed to protect critical infrastructures from evolving cyber threats. This study proposes a hybrid IDS framework that combines Graph Attention Networks (GAT) and Bidirectional Gated Recurrent Units (BiGRU) for privacy‑preserving distributed detection. The model is optimized with the Grey Wolf Optimizer (GWO) and enhanced through Federated Learning (FL). In IIoT traffic, GAT captures complex structural links, while BiGRU analyzes bidirectional temporal patterns, enabling accurate anomaly detection. GWO automates hyperparameter tuning and offers faster convergence than traditional methods such as Ant Colony Optimization. FL trains models locally on distributed IIoT devices, preserving data privacy and supporting decentralized deployment. The framework demonstrates improved scalability potential through decentralized training and reduced communication overhead (20% lower in a 10-node simulation), achieving detection accuracies of up to 95% across diverse attack scenarios, including Distributed Denial of Service (DDoS), Advanced Persistent Threats (APTs), and Zero‑Day exploits. It has been evaluated on the Edge‑Industrial Internet of Things dataset (Edge‑IIoTset), Canadian Institute for Cybersecurity – Internet of Things 2023 Dataset (CICIoT2023), and Real‑Time Internet of Things 2022 (RT‑IoT2022) datasets. An Explainable AI (XAI) module further improves interpretability by leveraging GAT’s attention mechanism. Overall, this technology demonstrates competitive offline performance on the EDGE-IIoTset, CICIoT2023, and RT-IoT2022 benchmark datasets, achieving F1-scores of up to 0.94, and shows promising scalability potential through decentralized Federated Learning with 20% lower communication overhead in a 10-node simulation. However, inference latency on resource-constrained edge hardware remains a challenge (e.g., 120–180 ms per sample on Raspberry Pi 4), which limits its strict real-time feasibility in mission-critical environments. Therefore, further model compression, adversarial robustness testing, and real-world deployment validation are required before practical edge-level applicability can be confirmed.

Federated Learning (FL) enables collaborative model training across distributed healthcare institutions without requiring the exchange of sensitive patient data. However, existing FL systems face challenges such as high communication overhead, data heterogeneity, and limited scalability. Additionally, the reliance on a single global model restricts the ability to effectively process multi-modal healthcare data. This paper proposes a cluster-based multi-model federated learning framework integrated with knowledge distillation. Clients are grouped based on data similarity, and specialized models are assigned to each cluster. Model outputs are fused at the central server, and knowledge distillation is used to compress multiple models into a lightweight global model. The framework is designed within an edge-cloud architecture to support scalable and practical deployment in real-world healthcare systems.