Federated Learning (FL) has emerged as an important paradigm for distributed intelligence in large-scale AI-based IoT systems under the Integrated Sensing, Memory, Communication, and Computation (SMCC) framework. However, its privacy guarantees may be threatened by gradient leakage attacks under dynamic training conditions. Existing client-side attacks often suffer from limited reconstruction fidelity and high computational overhead. In this paper, we propose PAFS, a generative gradient leakage framework based on Poisoning-driven Analytical Feature Separation, designed to investigate privacy vulnerabilities relevant to SMCC-enabled collaborative learning. PAFS introduces a feature separation mechanism in which strategically poisoned updates amplify target-class gradients, enabling target-related features to be extracted from fully connected (FC) layers. To bridge the semantic–spatial gap in reconstruction, we further introduce a Hierarchical Feature Fidelity (HFF) mechanism that enforces multi-scale feature consistency to preserve both fine-grained textures and semantic structure. As a result, target images can be reconstructed via a single forward pass. Experiments on CIFAR100 and ImageNet show that PAFS achieves competitive reconstruction quality with improved efficiency compared with representative baselines, while maintaining stable performance under varying client scales and Byzantine-robust aggregation rules, and remaining effective under low-magnitude gradient perturbations. The results highlight potential privacy risks in SMCC-enabled FL systems and provide insights for privacy evaluation in intelligent IoT systems.
The YOLO series has achieved remarkable progress in real-time object detection, yet nano-scale variants remain constrained by single-round feature pyramid topologies where the global correlation prior is distributed only once and extreme pyramid levels exchange information indirectly. In this paper, we propose YOLO-Prism, an efficient nano-scale detector built upon YOLOv13 that addresses this single-round limitation through asymmetric multi-round refinement. We propose a Multi-Round Feature Pyramid Network (MR-FPN) that exploits the decoupling between YOLOv13’s precomputed correlation prior and its lightweight gated distribution mechanism, appending a secondary top-down refinement pass at only ∼6% parameter overhead under a “compute once, distribute many” paradigm. Subsequently, as a near-zero-cost complement, we selectively integrate DCT-based Multi-Spectral Channel Attention (MSCA) at gradient-identified critical layers, contributing an additional gain at merely 0.01M parameters. Extensive experiments on MS COCO, PAS-CAL VOC, and VisDrone benchmarks demonstrate that YOLO-xsPrism achieves 41.66% AP50:95 on COCO val2017, surpassing the strong YOLOv13n baseline by 0.62 points under identical training conditions and outperforming YOLOv12n by 1.10 points, at 2.63M parameters and 7.1 GFLOPs. On YOLOv11n, whose neck lacks such a prior, the same asymmetric topology yields 0.50 AP50:95 at 3.8% overhead, confirming that the benefit of multi-round refinement is independent of any specific gating mechanism. This lightweight algorithm can be applied to intelligent video measurement, industrial edge detection sensors, and IoT monitoring.
Dynamic metasurface antennas (DMAs) are emerging as a promising technology for future satellite communications, offering reductions in power consumption and hardware costs. However, the proliferation of Internet of Things (IoT) applications has exacerbated spectrum scarcity in DMA-assisted satellite communication systems. To tackle this limitation, we propose a novel integrated reconfigurable intelligent surface (RIS) and DMA empowered satellite symbiotic radio (SR). Specifically, the satellite equipped with a DMA transmits signals to primary users with the assistance of the RIS. Meanwhile, following the RIS-assisted SR principle, the RIS transmits its own signal to the secondary user. Considering low-resolution digital-to-analog converters (DACs) at the transmitter, we investigate the maximization of the weighted sum rate (WSR) through the joint optimization of the transmit beamforming vectors, the weight matrix associated with the DMA, and the phase shifts of the RIS. The strong coupling of optimization variables and the structural limitations inherent to the DMA make the problem highly challenging. To overcome these difficulties, we develop an efficient alternating optimization method based on the Lagrangian dual transform and quadratic transform algorithms to reformulate the original problem as a computationally tractable form. Then, we apply the penalty convex-concave procedure principle and the complex circle manifold algorithm to design the RIS phase shifts and the DMA weight matrix, respectively. In addition, an energy efficiency (EE) maximization formulation is developed to further evaluate the power-consumption benefit of the DMA-assisted transmitter. Numerical results demonstrate the effectiveness of the developed algorithm and highlight the favorable rate-power tradeoff of the DMA-assisted architecture, which achieves competitive WSR with improved EE compared with the conventional full-digital scheme.
With the emergence of the Internet of Things (IoT), the proliferation of interconnected smart devices is reshaping ordinary environments into smart spaces. This transformation creates new requirements for user authentication systems that operate transparently and continuously to ensure both security and usability in IoT environments. Leveraging the ubiquity of existing wireless infrastructure, WiFi-based authentication could enable unobtrusive, real-time user identification without requiring active participation. These advantages position WiFi-based systems as a promising solution for seamless, user-transparent, and continuous authentication. Building on this premise, this survey offers a comprehensive review of existing WiFi-based authentication systems and introduces a unified four-layered framework. Additionally, we outline key challenges and future research directions, aiming to unlock further advancements in WiFi-based authentication systems to match the increasing complexity of smart spaces.
Gesture-aware wearable interfaces for human-centric Internet-of-Things (IoT) systems require compact sensing front ends that can provide finger-level observability without dense wiring, bulky sensor arrays, high-dimensional RF measurements, or computation-heavy learning models. This work presents a textile guided-wave RF sensing method that repurposes a glove-integrated 1×4 Wilkinson power divider as a four-channel finger-posture transducer at 865MHz. Unlike its conventional role as a passive RF power-distribution network, each output branch is routed along an individual finger region so that extension/flexion perturbs the local guided-wave propagation path and modulates the corresponding input-to-branch transmission magnitude. The structure, fabricated on a flexible cotton substrate, provides four synchronized branch-magnitude observables, |S21|, |S31|, |S41|, and |S51|, associated with Ports 2–5. The branch responses are acquired using software-defined radios and processed through a deterministic edge-processing pipeline comprising signal conditioning, denoising, and adaptive-hysteresis digitization to generate compact binary finger states. An application-level lookup rule then maps the resulting 4-bit state vector to commands for live dashboard visualization. Measurements confirm good multiport matching, with all port reflection coefficients below −10 dB at 865MHz, near-balanced input-to-branch transmission responses, and repeatable posture-dependent magnitude modulation. The extracted branch responses are subsequently converted into representative per-finger binary states and streamed as compact state packets for real-time monitoring. The results support the divider-as-sensor principle as a compact proof-of-concept route toward textile-integrated IoT wearables for gesture-aware HMI, HRI, and connected control applications.
The digitalization of high-voltage substations—driven by the IEC 61850 standard—has significantly expanded the cyber-attack surface, exposing protection systems to increasingly sophisticated threats. While prior research has examined individual protection functions or attack modeling in isolation, a comprehensive vulnerability analysis of substations and their integrated protection functions remains a critical gap. This paper addresses this gap through the first systematic analysis of vulnerabilities across all protection functions for each major substation relay type, considering cyber-attack paths, exploitation techniques, and defense mechanisms encompassing network-based, physics-based, and hybrid approaches. Adopting both component-centric and function-centric perspectives, the analysis examines vulnerabilities at the protection relay level and within their functional operations. Additionally, to demonstrate practical implications, a real-world substation is simulated in PSCAD/EMTDC, evaluating two cyber-attack scenarios: (i) targeting distance protection through sampled value (SV) false data injection and (ii) compromising communication-assisted protection via GOOSE message poisoning. The results show that coordinated manipulation of SVs, even with minimal perturbations, can induce false trips or prevent correct protection operation while remaining difficult to detect. In addition, poisoning GOOSE messages can cause delayed or incorrect tripping of otherwise healthy transmission lines. Collectively, these findings highlight the urgent need for physics-based defense mechanisms that extend beyond traditional network-layer security measures.
Smart-home trigger-action programming (TAP) allows users to compose automation rules, yet rule interactions can produce device clashes, environmental conflicts, trigger chains, and safety or privacy violations. Existing detectors either require manually maintained device semantics or produce opaque decisions. This paper introduces CAGE-TAP, a concept-augmented graph-evidence framework that routes every conflict prediction through a compact vector of human-checkable TAP concepts—structural concepts computed from normalized rule fields and semantic/risk concepts learned from pair-centered interaction graphs—so that explanations are faithful to the implemented decision path and correctable at inference time. We evaluate CAGE-TAP on 21,447 labelled rule pairs obtained by aligning an audited 580-pair collection TAPConflict-580 (TC-580), with IoTCOM and TAPFixer interactions. The benchmark contains 10,072 conflicts and 11,375 non-conflicts under group-disjoint train, validation, and test partitions. CAGE-TAP with an R-GCN encoder achieves 0.9280 ± 0.0004 binary F1, 0.9519 ± 0.0004 AUPRC, and 0.9020 ± 0.0035 seven-type Macro-F1. Its type diagnosis exceeds a capacity-matched direct R-GCN (0.7863 ± 0.0011) and fine-tuned BERT (0.8030±0.0049). The six learned semantic/risk concepts reach 0.8532±0.0078 Macro-F1. Targeted contribution masking changes twice as many predictions as matched random masking, while reference correction improves both binary and typed performance. Each positive conflict type is represented by at least 72 evaluated test instances; per-type F1 ranges from 0.8221 to 0.9467.
Joint detection and communication (JDC) reduces hardware footprint and power consumption through resource sharing, which is promising for the underwater Internet of Things (UIoT). Meanwhile, advances in reconnaissance technologies pose increasingly stringent covertness challenges to UIoT. This paper presents a bio-inspired covert JDC (BCJDC) method that exploits dolphin clicks for covert transmission in UIoT environments. The phase features of dolphin clicks are extracted using the amplitude-weighted phase-locking value (awPLV) to convey communication information. In addition, the inter-click intervals (ICIs) of the click trains are designed based on Golomb rulers, in which the distances between any two marks are unique, to effectively suppress autocorrelation function (ACF) sidelobes. Experimental results show that the proposed PMM schemes achieve communication data rates ranging from approximately 0.8 kbps to more than 2.5 kbps, with BERs ranging from 0 to below 0.1% across tank and sea experiments. Target detection was also demonstrated in tank experiments and in sea experiments at transmission distances of 16.6 m and 66.0 m, while the tank results maintained centimeter-level ranging accuracy. Simulation results, tank experiments, and sea experiments verify the efficacy of the proposed method for covert detection and communication for UIoT applications.
Global navigation satellite system (GNSS) precise point positioning (PPP) provides high-accuracy positions for vehicle-borne mobile Internet of Things (IoT) platforms without local reference stations, enabling georeferencing of sensing data for urban mobile mapping and road-asset inspection. However, dense urban environments introduce frequent signal blockage, cycle slips, and multipath effects, degrading PPP accuracy and availability. This paper proposes a time-differenced carrier phase (TDCP)-aided PPP framework. A robust Mahalanobis-type distance is employed for TDCP quality control to down-weight anomalous observations in velocity estimation. TDCP post-fit residuals are used to identify cycle slips and multipath-contaminated observations. Cycle slips are handled through ambiguity reinitialization, whereas multipath effects are mitigated by inflating ambiguity process noise. TDCP-derived velocity and its covariance are incorporated into the PPP filter to strengthen state propagation. Experiments using two urban vehicular datasets show that the quality-control strategy yields velocity root mean square (RMS) values below 2 cm/s in most cases. Compared with TurboEdit, the TDCP-based cycle slip detection method reduces false alarms and missed detections by up to 99.7% and 89.8%, respectively. The multipath mitigation strategy reduces carrier phase post-fit residuals and improves the correct ambiguity fixing rate. The complete framework increases the proportions of epochs with horizontal positioning errors below 0.1 m to 94.6% and 55.8% for the two datasets, respectively, and those with vertical positioning errors below 0.2 m to 89.9% and 65.0%. The corresponding East/North/Up RMS errors are reduced to 3.7/5.6/13.2 cm and 9.2/16.6/29.8 cm, respectively. These results indicate that the proposed framework enhances GNSS-based positioning for urban mobile IoT platforms.
This paper aims to design a decentralized controller for large-scale systems under denial-of-service (DoS) attacks by utilizing exclusively open-loop operational data without requiring any system parameters. Firstly, we establish a Lyapunov function with the descriptor method to develop model-based stability conditions for large-scale systems under DoS attacks. Then we employ a bounded noise signal to stimulate the system and gather the system state and input data that satisfy some requirements. By integrating this data with the model-based stability conditions, we employ the full block S-procedure to obtain a data-based design criterion. In addition, the derived data-based results are extended to the large-scale system with state-delays by using appropriate Lyapunov-Krasovskii (L-K) functional. Finally, the effectiveness of the presented method is illustrated by two examples including a platoon of heavy-duty vehicles (HDVs).
Internet of Drones (IoD) systems increasingly support mission-critical applications that require secure collaboration, selective data access, and privacy-preserving analytics under stringent resource and latency constraints. Existing solutions typically address authentication, encrypted storage, searchable retrieval, or blockchain-based trust independently, without providing an integrated framework for authorized encrypted search, verifiable retrieval, and post-quantum resilience. This paper presents a lightweight, post-quantum-aware authentication and data-access framework for fog-assisted IoD based on a unified epoch-driven architecture. Drone identities are established through PUF-sealed mutual authentication with ML-KEM, eliminating persistent private-key storage. The framework introduces Capability-Bound Searchable Objects (CBSOs), Authorization-Bound Trapdoors (ABTs), and a Two-Level Capability-Aware Search (TCAS) mechanism to enable policy-isolated encrypted retrieval without online policy evaluation. Retrieval integrity is ensured through Verification-Ready Retrieval Packages (VR-RPs), validated using Sparse Merkle proofs against blockchain-anchored searchable-state commitments. Experimental results show that the proposed framework incurs only a 10% authentication overhead during the first epoch exchange and becomes 45% faster than classical RSA-based authentication thereafter. Compared with representative schemes, it achieves up to 95% faster trapdoor generation, 90.6% faster encrypted search, 8.5× higher verifiable retrieval throughput, and reduces drone-side computation and energy consumption by up to 98.6% and 97.9%, respectively. Formal security analysis demonstrates authentication soundness, forward-secure epoch isolation, authorization correctness, adaptive searchable privacy, verifiable retrieval, and resistance against replay, trapdoor-forgery, Sybil, flooding, and insider-fog attacks, demonstrating the practicality of the framework for large-scale resource-constrained IoD deployments.
Internet of Medical Things (IoMT) infrastructures can distribute pathology–genomics processing, yet no sufficiently powered public cohort links pretreatment cervical whole-slide images (WSIs), tumor RNA, and standardized immune-checkpoint-inhibitor outcomes. We present CERVITWIN, an IoT-enabled prespecified edge–cloud protocol. Hospital-edge processing converts quality-controlled tiles into 64 morphology prototypes while retaining raw images locally; a fixed pathway matrix produces 50 genomic tokens. A perturbation-stable sparse graph and reliability masks support incomplete-modality fusion. Heterogeneous response cohorts supervise a genomic teacher, but cross-domain distillation remains hypothesis-generating until a joint cervical cohort is available. Disjoint temperature and conformal-calibration sets govern prediction and deferral. TCGA-CESC, TIGER, and GSE205247 receive nonoverlapping analytical roles. Four protocol-verification experiments retain the reported numerical results for score, stress/calibration, graph-export, and payload-accounting checks; they are not target-domain clinical estimates or measured device/network results. The protocol fixes tensor dimensions, losses, provenance, endpoint, fault, security, and external-test rules. Clinical and deployed-IoT claims require frozen multi-institution data and repeated hardware, network, energy, and privacy measurements.
Time-sensitive IoT applications require forecasts that are both accurate and timely. However, existing spatiotemporal forecasting methods usually treat prediction time as a fixed observation horizon shared by all nodes, which is fundamentally mismatched with the asynchronous emergence of predictive signals across locations. We formulate adaptive early spatiotemporal forecasting as a coupled node-wise decision problem, where the model must determine when to predict and what to predict from partial observations under evolving spatial dependencies. To this end, we propose ESTMGCN, a unified framework that learns node-wise prediction timing while preserving informative spatiotemporal dependencies and reducing error accumulation in early forecasting. Experiments on four real-world datasets show that ESTMGCN consistently outperforms strong baselines under different target observation budgets and achieves a better trade-off between timeliness and accuracy.
This paper proposes a gray prediction and fuzzy logic-based handover (GPFLH) algorithm to address handover challenges arising from highly dynamic network topology and inherent heterogeneity in satellite-terrestrial integrated networks. We construct a comprehensive system model that captures key features including the high-speed movement of low earth orbit satellites, ground user mobility patterns, and wireless channel characteristics. Based upon this model, we design an adaptive handover triggering mechanism utilizing gray prediction. Specifically, this mechanism employs the GM(1,1) model to perform short-term trend forecasting of the signal-to-noise ratio (SNR) and dynamically adjusts the handover trigger advance time, thereby compensating for substantial transmission delays in satellite-terrestrial links and mitigating the risk of decision latency. Furthermore, we establish a multi-attribute fuzzy decision-making framework that comprehensively evaluates three key attributes, namely SNR, remaining visible time, and remaining network load. Through fuzzification and quantitative evaluation, this framework achieves intelligent trade-offs among multiple criteria and enables optimal handover target selection. Extensive simulation results demonstrate that, compared to existing benchmarks, the proposed GPFLH algorithm effectively reduces handover frequency, suppresses ping-pong handovers, minimizes unnecessary handovers. Under favorable channel conditions, it also achieves superior communication performance compared to existing benchmarks. Therefore, the proposed GPFLH algorithm provides an effective solution for achieving effective and seamless network coverage.
The precision of positioning information provided by global navigation satellite system (GNSS) is important for Internet of Things applications in smart cities. Multipath effects, mainly including multipath interference and non-line-of-sight (NLOS) reception, are the main errors that affect the positioning precision. The multipath hemispherical map (MHM) mitigates multipath interference, but NLOS reception is often ignored. In this study, we propose the refined multipath and NLOS hemispherical maps (MNHMs) to mitigate both NLOS reception and multipath interference simultaneously. Specifically, first, NLOS reception and multipath interference are preliminarily separated using the carrier-to-noise density ratio (C/N0) nominal function, and then the NLOS hemispherical map (NHM) is established for the first time, and the MHM is also established; Second, the refined MNHMs are constructed by adjusting and interpolating the grid correction values of MHM and NHM to improve the accuracy and availability; Third, the NLOS reception and multipath interference are corrected simultaneously. To validate the proposed method, two consecutive 48-hour RTK occlusion positioning experiments were carried out. The positioning results show that the positioning precision of the proposed method is improved compared with that of the uncorrected and conventional methods based on MHM. Specifically, the proposed method achieves average reductions of 23.5% and 29.5% in the 2D and 3D root mean square errors, respectively. Meanwhile, the residuals of both datasets show a significant reduction, indicating that the proposed method effectively mitigates multipath and NLOS errors. In addition, a practical observation dataset is used to verify the reliability of the proposed method.
Dynamic risk assessment in esophageal squamous cell carcinoma (ESCC) can draw on radiology, pathology, omics, and clinical records, although these data rarely arrive together or remain complete. We developed IoT-DTFormer as a proof-of-concept IoT interface and learning architecture for updating a persistent patient representation from asynchronous multimodal observations. Separate encoders represent tumor–lymph-node geometry, hierarchical pathology, pathway-guided omics, and clinical records; reliability-aware fusion combines availability, a learned gating score, and observation age. The internal evaluation used 90 patients with 30 events, but CT was available for only 13. At the prespecified baseline landmark, the five-fold mean C-index was 0.78; this is a descriptive cross-validation estimate without a patient-level confidence interval and is not evidence of generalizable or comparative performance. All 90 patients contributed one evaluable risk state, so the dynamic update mechanism was not longitudinally validated. GSE53624 and GSE53622 were analyzed by within-component cross-validation of the omics branch and do not constitute external validation of the complete architecture. Device measurements began after token delivery and excluded acquisition, communication, conversion, security, queuing, and hospital-system integration. The findings therefore support only technical feasibility and motivate prospective multicenter longitudinal evaluation.
The Internet of Underwater Things (IoUT) extends distributed sensing to marine environments, where unmanned underwater vehicles (UUVs) serve as mobile nodes for flexible underwater operations. This article investigates the heterogeneous UUV multi-type task planning problem (HUMTTP) for a collaborative system comprising autonomous underwater vehicles (AUVs), underwater gliders (UGs), and bionic manta-ray underwater vehicles (BMUVs). Existing methods generally prioritize execution efficiency but often overlook platform heterogeneity and acoustic communication limitations. To address these issues, this article proposes a hierarchical task planning framework called REASOM-UCLNS. In the task allocation layer, the reward-and energy-aware self-organizing map (REASOM) algorithm incorporates a novel neuron-based reward estimation strategy into winner selection. By integrating the estimated platform-specific task rewards with current-aware energy evaluation, REASOM assesses the execution suitability of heterogeneous UUVs and facilitates the efficient allocation of multi-type tasks. In the route planning layer, the underwater cooperative large neighborhood search (UCLNS) algorithm converts the generated task sets into feasible routes for individual UUVs. UCLNS develops two constraint-specific operators to restore route feasibility when communication connectivity or energy balance violations occur. By integrating targeted constraint handling into the destroy-repair process, UCLNS improves route quality while satisfying operational constraints. Numerical results demonstrate the superior performance of the proposed framework. Finally, a lake experiment involving four heterogeneous AUVs further confirms its practical applicability.
