Federated Anomaly Detection for IoMT Networks: Privacy‐Preserving Design, Lightweight Implementation, and Runtime Evaluation

An Intelligent Intrusion Detection and Privacy-Preserving Architecture for the Internet of Medical Things (IoMT)

The Internet of Medical Things (IoMT) is transforming healthcare delivery, but brings significant security and privacy challenges due to the diverse range of devices, sensitive patient data, and real-time operation requirements. Existing Intrusion Detection Systems (IDS) have improved detection and privacy through approaches such as federated learning and blockchain, yet they focus primarily on network-level attacks, overlooking device-level attacks which is the primary source of data leakage, device unavailability, and model poisoning in federated learning (FL)-based approaches. For instance, Bring Your Own Device (BYOD) introduces heterogeneity and Non-Independent and Identically Distributed data (non-IID) distributions that affect the performance of conventional FL approaches. We therefore propose an intelligent, lightweight Tiny LSTM–GRU hybrid IDS on the edge to monitor device-generated behavioral patterns in real time, with minimal computational and energy overhead. To preserve privacy and handle the issue of non-IID data across heterogeneous IoMT devices, we propose an adaptive FedProx-based weighted federated learning framework. Our proposed framework achieves an overall accuracy of 99.1% on the edge with latency between 1.8ms per sample, with a mean global accuracy of 99.4% and global loss of 0.037 during convergence, making it highly suitable for real-world IoMT deployments.

Towards Privacy-Preserving and Continual Intrusion Detection Systems in Internet of Medical Things

The rapid expansion of the Internet of Medical Things (IoMT) introduces critical vulnerabilities into healthcare infrastructures, demanding the development of enhanced cybersecurity strategies. This domain faces significant challenges arising from emerging and previously unknown attacks, combined with the decentralized nature of network traffic data, which is rarely shared due to stringent privacy constraints. To address these challenges, we propose a Federated Class Incremental Learning (FCIL) framework for Network Intrusion Detection System (NIDS) in the IoMT domain, enabling the continuous recognition of new attacks leveraging distributed data without exposing sensitive information. Findings on the CIC-IoMT24 dataset show that combining the incremental approach $\text{BiC}^{+}$ with the federated aggregation algorithm FedDyn achieves 71% F1 score in fine-grained misuse detection. While exhibiting a 14% F1 drop compared to the ideal centralized training-from-scratch, our approach ensures data privacy and computational efficiency.

Federated Learning for Privacy-Preserving Anomaly Detection in Heterogeneous IoT Networks

The rapid proliferation of Internet of Things (IoT) devices across smart homes, industrial plants, healthcare systems, and wearable platforms has generated an unprecedented volume of distributed, heterogeneous, and privacy-sensitive data. Conventional centralized anomaly detection pipelines require raw sensor and network traffic data to be transmitted to a central server, exposing sensitive information to interception, misuse, and regulatory non-compliance while also incurring substantial communication overhead. This paper proposes a Federated Learning (FL) framework for privacy-preserving anomaly detection tailored to heterogeneous IoT networks characterised by non-independent and identically distributed (non-IID) data, variable computational capacities, and intermittent connectivity. The proposed architecture couples a lightweight convolutional-recurrent local model with a differential-privacy-augmented Federated Averaging (FedAvg) aggregation strategy and an optional secure-aggregation layer to prevent gradient leakage. A dynamic client-selection and adaptive-weighting mechanism mitigates statistical heterogeneity across device clusters, while a local outlier-scoring module enables edge-level anomaly flagging without exposing raw payloads. Extensive simulation on merged benchmark traffic derived from N-BaIoT and CICIDS2017 characteristics, partitioned across simulated smart-home, industrial, and wearable clusters, demonstrates that the proposed framework attains 97.4%–97.7% detection accuracy, within one to two percentage points of a fully centralized, non-private baseline (98.6%), while reducing raw-data transmission to zero and lowering communication overhead relative to naive parameter exchange. Comparative results against local-only training confirm that federated collaboration yields a 9–10 percentage-point accuracy improvement under severe data heterogeneity. These findings indicate that the proposed method offers a practical, scalable, and regulation-compliant pathway toward trustworthy intrusion and anomaly detection in large-scale, heterogeneous IoT deployments.

Federated Learning and LLM-Driven Threat Intelligence for Zero Trust IoT Architecture

While the Internet of Things (IoT) has become essential, they introduced serious security and privacy challenges, especially for mission-critical environments. Legacy devices are vulnerable to viruses, data breaches, and unauthorized access, and updating these devices would be infeasibly costly. As a solution, this paper presents a Federated Learning and LLM-Driven Threat Intelligence for Zero Trust IoT Architecture, with FL for anomaly detection integrating privacy-preserving distributed learning, continuous identity verification, and LLM-driven autonomous threat response into a unified pipeline. Unlike existing solutions, our framework enforces Zero Trust at every communication layer via mutual TLS (mTLS) over MQTT, ensuring no device or message is implicitly trusted. Our experiments with Raspberry Pis and various sensors achieve an F1 score of 0.9091 and an ROC-AUC of 1.0, highlighting the effectiveness of the proposed framework in enabling privacy-preserving anomaly detection for resource-constrained IoT devices.

Fed-blam: federated BERT and LLaMA for IoT malware detection

Federated TinyML and digital twin framework for secure and resilient IoMT-based ICU monitoring.

Resource-constrained medical sensing devices are increasingly expected to support local intelligence, privacy-preserving collaboration, and secure communication in Internet of Medical Things (IoMT) environments. However, deploying federated learning in ICU monitoring remains challenging because hospital data are often non-IID, model updates may be adversarially poisoned, and emerging quantum-security threats require stronger communication protection. This paper presents a federated TinyML framework with an edge-hosted patient-state Digital Twin layer for ICU monitoring. The Digital Twin component is implemented as a lightweight patient-state representation rather than a full physiological simulator. It maintains recent physiological observations, temporal risk trends, predicted status labels, and interpretable alert information at the hospital gateway. Patient-specific adaptation is therefore achieved through individualized temporal state tracking and risk-history synchronization, while the predictive model is collaboratively learned across hospitals. The suggested framework employs decision-tree ensembles due to their lightweight nature, efficient inference capabilities, and interpretability inherent to ESP32-class devices. As decision tree structures are not amenable to average aggregation like neural network parameters, the aggregation of local client models is done via the validation-based ensemble fusion. In order to enhance the resistance of the framework to attacks based on poisoning clients, the Performance-Based Filtering (PBF) method analyzes the performance of each local ensemble against a trusted validation set and excludes low-quality local ensembles from the subsequent global ensemble aggregation. The secure model exchange process utilizes ML-KEM-512 key encapsulation scheme in combination with AES-256-GCM authenticated encryption. The extra overhead of the crypto layer is 0.09 ms per update exchange. The framework is evaluated using a clinician-reviewed synthetic ICU monitoring dataset distributed across three non-IID hospital clients and externally benchmarked using a PhysioNet-based critical-care dataset constructed from shared physiological features. The prediction task uses five status categories: normal, mild, moderate, critical, and an outlier/anomaly category; these labels represent synthetic severity-status classes rather than prospectively validated ICU endpoints. Under a targeted label-flipping attack in which one client relabels critical cases as normal, the proposed PBF strategy achieves 86.79% accuracy, 0.782 macro-F1, 0.904 critical-class recall, 0.837 critical-class F1-score, 0.947 macro-AUROC, and 0.799 macro-AUPRC on the synthetic ICU evaluation. External benchmarking at [Formula: see text] achieves 98.04% accuracy, 0.959 macro-F1, 0.977 critical-class recall, 0.987 critical-class F1-score, 0.999 macro-AUROC, and 0.983 macro-AUPRC. Comparative evaluation against FedAvg, Krum, Trimmed Mean, and Coordinate Median shows that PBF provides stronger protection of critical-class recall under adversarial conditions, although full prospective clinical validation remains outside the scope of this study. Overall, the results indicate that a federated TinyML architecture with lightweight patient-state tracking, validation-based ensemble filtering, differential privacy, and post-quantum-secure communication can support privacy-aware and attack-resilient ICU monitoring experiments in resource-constrained IoMT settings. The study should be interpreted as a technical feasibility and robustness evaluation rather than a fully integrated clinical Digital Twin deployment.