Search Results (576)

The deployment of 5G standalone (SA) networks introduces cloud-native core architectures, service-based interfaces, and programmable radio access networks that substantially expand the mobile attack surface. Existing work has focused mainly on protocol-level vulnerabilities or isolated anomaly detection, with less attention to SOC-level monitoring, correlation, and investigation of 5G-specific threats. This paper presents a SOC-integrated, protocol-aware detection framework for 5G SA environments, combining structured telemetry ingestion, log normalization, decision-based detection logic, and analyst-oriented visualization within an Elastic-Stack-based SOC architecture. The approach is implemented on an experimental 5G SA testbed at Obuda University and evaluated using a dual-source design: controlled testbed scenarios complemented by an observational analysis of telemetry from an independent, large-scale international cyber-defense exercise whose adversarial complexity is difficult to reproduce synthetically. The evaluation exercises attack scenarios including rogue network element registration, authentication abuse, and control- and user-plane manipulation. The results indicate that embedding 5G-aware detection logic into SOC workflows supports situational awareness and structured incident investigation. A single-layer versus cross-layer ablation of the published detection rules quantifies where cross-layer correlation is strictly required for detection versus where it primarily enriches interpretation. This is extended with a limited generic-SIEM-style baseline check and a live, a priori repeated-trial sensitivity check ($N=30$, Wilson 95% CI [88.6%, 100%]). The framework offers a reproducible methodological foundation for operational 5G security monitoring and practical guidance for next-generation mobile network defense. Future work will extend controlled repeated-trial evaluation to the remaining detection rules, broaden the SIEM baseline comparison, and assess generalizability beyond Open5GS. Full article

Smart textile interfaces hold promise for battery-free wearable interaction, yet their adoption is limited by complex fabrication and insufficient on-body evaluation. We present TriboTex, a low-cost spray-patterning workflow that forms nylon–Cu–nylon triboelectric stacks on cotton textiles using laser-cut PET stencils and commercially available materials. The core consumables cost approximately USD 0.003/cm², and sensor geometry can be rapidly iterated by modifying only the digital stencil. Controlled characterization across nine devices from three fabrication batches showed a peak open-circuit voltage of 52.3 V and a maximum power density of 1870 µW/m² at 4 GΩ. The output retained 96.1% of its initial voltage after 1000 bending cycles and 94.2% after 24 h of simplified saline immersion. Three-sample environmental sweeps showed voltage amplitudes of 41.9–43.7 V from 15 to 45 °C, with a decrease to 27.7 V at 0 °C; the humidity response remained within 92.7–104.5% of the 20% RH value over 20–60% RH but decreased to 19.9% at 70% RH. Two wearable prototypes were developed: a single-electrode garment sleeve recognized tap, double-tap, and swipe gestures with 95.0% accuracy across 1200 trials from 12 participants; a single-electrode insole generated action-dependent peak voltages up to 123 V under repeated foot loading and was connected through a rectification and voltage-regulation module to charge a battery. Across the two 12-participant studies, attachment and fit stability emerged as shared integration requirements, while participant feedback and controlled humidity measurements highlighted moisture management as a priority for reliable on-body sensing and energy capture. The primary contribution is an accessible, low-cost, and geometry-flexible route for early-stage wearable sensing experiments and application demonstrations, supported by documented fabrication, electrical characterization, and human-centered evaluation. Full article

Log parsing converts raw system logs into structured templates and is a key preprocessing step for Artificial Intelligence for IT Operations (AIOps). Existing parsers face a practical trade-off: rule-based methods offer high throughput but limited adaptability across heterogeneous log sources, whereas Large Language Model (LLM)-based parsers achieve broader semantic coverage at the cost of inference latency, privacy exposure, and cloud dependency. This paper presents FlanBC, a log parsing framework that formulates template extraction as a BIO (Beginning, Inside, Outside) sequence-labeling task and integrates a Flan-T5 semantic encoder, Bidirectional Long Short-Term Memory (BiLSTM) layers for local sequential modeling, and a Conditional Random Field (CRF) decoder for structured label prediction. Log-specific preprocessing and a subword-to-token alignment mechanism adapt the general-purpose encoder to semi-structured log data. A layer-freezing strategy reduces the number of parameters updated during training. The framework supports local inference without external API dependency. Experiments on three benchmark datasets from LogHub (HDFS, BGL, OpenStack) under a supervised random-split setup evaluate parsing accuracy, training efficiency, statistical stability across random seeds, and component contributions. FlanBC achieves a Group Accuracy of 99.32% on HDFS and 98.47% on BGL, with an inference throughput of 700+ logs/s on a consumer-grade GPU. On OpenStack, performance is lower (GA = 92.54%), reflecting the challenge that diverse natural-language-like logs pose for compact encoder-based models. Under a stricter template-disjoint split that prevents template overlap between training and test sets, FlanBC achieves an average Group Accuracy of 91.14%, indicating that the model generalizes to unseen templates beyond in-distribution recognition. Ablation results indicate that the semantic encoder, BiLSTM module, and CRF decoder each contribute to prediction accuracy. These findings suggest that domain-adapted semantic encoders combined with structured decoding offer a practical accuracy–efficiency balance for log parsing in settings where local, cloud-free inference is preferred. Full article

Water scarcity in semi-arid regions makes efficient irrigation scheduling a priority, yet farm-level adoption of soil-moisture monitoring remains limited by the cost, low portability and installation complexity of commercial sensing systems. This study presents the development and field validation of WaziSense, a low-cost, solar-powered Internet-of-Things (IoT) smart tensiometer, developed within the OSIRRIS platform for soil-water monitoring and irrigation scheduling. The device couples a Watermark granular-matrix sensor and a DS18B20 temperature probe to an ATmega328P microcontroller (Arduino Pro-Mini, 3.3 V, 8 MHz) with long-range LoRa communication and a maximum-power-point-tracking (MPPT) solar-charging stage, logging soil matric potential and soil temperature every 15 min. An open-source edge/cloud stack (WaziGate, WaziApp) retrieves weather forecasts from an open API and runs an automated machine learning (AutoML) regression pipeline that forecasts soil-water dynamics and the time to a user-defined threshold, from which irrigation is scheduled and its applied volume verified by a flow meter. The system was deployed at three bioclimatic sites in Tunisia (durum wheat at Cherfech, citrus at Nabeul, apple at Sbeitla), with tensiometers installed at 20 and 40 cm depths, and validated against commercial 10HS capacitive probes coupled to a ZL6 data logger, with which the co-located readings were significantly correlated (r = 0.81). Calibrated readings showed a strong relationship between soil–water content and soil–water potential (R² = 0.99), and the edge forecasting model reproduced soil–water dynamics on unseen data (Sbeitla apple site, 5-day horizon) with R² = 0.73, RMSE = 0.35, MAE = 0.23 and MPE = 12.52%. With a material cost under about 90 EUR per node and fully open-source hardware and software, WaziSense is one to two orders of magnitude cheaper than commercial monitoring stations, offering an affordable, reproducible and scalable tool for data-driven irrigation in water-limited agriculture. Full article

Deep Cretaceous clastic reservoirs in the Kuqa Depression are major targets for natural gas exploration in the Tarim Basin. Recent exploration of the Yageliemu Formation in the Ketan area has revealed considerable resource potential, although reservoir performance is strongly affected by deep burial, compaction, repeated fluid–rock interaction, and pronounced pore-system heterogeneity. Core descriptions, epoxy-impregnated thin sections, cathodoluminescence, scanning electron microscopy, conventional petrophysical measurements, mercury intrusion capillary pressure, and nuclear magnetic resonance data were integrated to evaluate sedimentary fabric, reservoir-space types, pore-throat characteristics, and diagenetic modification. The succession was deposited mainly in a braided river delta plain setting and is dominated by medium sandstone, pebbly sandstone, and fine conglomerate. Overall reservoir quality is poor, with an average porosity of 3.4% and permeability commonly between 0.01 and 0.5 mD. MICP and NMR data distinguish four pore-structure types. From Type I to Type IV, average displacement pressure rises from 0.89 to 11.02 MPa, whereas median throat radius and movable-fluid porosity decline from 0.17 to 0.01 μm and from 2.00% to 0.72%, respectively. Residual intergranular pores constitute the main storage space, while feldspar- and lithic-fragment-dissolution pores provide additional local storage. Fractures contribute little pore volume but can markedly improve connectivity where they remain open or only weakly cemented. The present reservoir heterogeneity reflects the combined effects of sand-body stacking, sandstone–mudstone arrangement, fault-related fracturing, and multistage diagenesis. The most favorable intervals occur in thick, relatively clean stacked sand bodies where residual pores are preserved, dissolution pores remain connected to the throat network, and fractures have undergone limited late-stage filling. Full article

Transforming waste and agricultural-residue biomass into bioenergy is central to the circular bioeconomy, yet routing such heterogeneous residues to the right thermochemical pathway depends on the higher heating value (HHV), which is conventionally measured by slow, resource-intensive bomb calorimetry. Here, we present an explainable alternative that predicts HHV from inexpensive elemental inputs. We used a publicly archived compilation of 344 literature-reported biomass samples retrieved from an open data repository rather than assembled by the authors, including carbon (C), hydrogen (H), oxygen (O), nitrogen (N) and sulfur (S). Measured HHV was the target. The samples spanned woody, herbaceous and agricultural-residue biomass, and they were standardized through duplicate removal, consistency verification and outlier assessment. On these features, we developed a stacked hybrid model combining Random Forest, eXtreme Gradient Boosting and Artificial Neural Networks, which estimated the HHV with $R^2$ = 0.99, RMSE = 0.45 MJ/kg and MAE = 0.30 MJ/kg. SHAP and LIME analyses showed that carbon exerts the strongest positive influence on HHV, whereas oxygen contributes negatively, which is consistent with established thermochemical principles. Within the compositional range covered by the training data, and subject to the absence of external validation, the framework offers a fast and interpretable complement to bomb calorimetry for screening residue biomass. Full article

Autonomous field robots are increasingly used in precision agriculture for monitoring, phenotyping, spraying, and site-specific intervention, yet reliable autonomy remains difficult under vegetation occlusion, uneven terrain, variable illumination, and intermittent communications. This review provides a deployment-oriented synthesis of navigation and sensor-fusion methods for agricultural robots, with emphasis on what is practical under field conditions rather than only in laboratory settings. The literature was examined through a structured narrative-review workflow using Scopus, Web of Science, IEEE Xplore, ScienceDirect, SpringerLink, and related citation tracking, with primary emphasis on studies published between 2015 and 2025. We compare global, local, and hybrid planning methods; motion-control strategies such as PID, Pure Pursuit, and MPC; and localization pipelines that combine GNSS, IMU, LiDAR, cameras, odometry, and SLAM or Kalman-family fusion. Beyond algorithm summaries, the review links method selection to agricultural deployment constraints, including GNSS degradation, dynamic obstacles, compute limits, ROS 2 integration, time synchronization, and coordinate-frame management. The synthesis shows that no single stack is optimal across all crop systems: lightweight GNSS/IMU-based solutions remain attractive in structured open fields, whereas orchards, vineyards, and other occluded environments benefit more from tighter multi-sensor fusion and SLAM-supported localization. Finally, the review distills design guidance for sensing, planning, validation, and digital-twin-supported testing, and identifies research gaps related to robustness, benchmarking, safety, and scalable on-farm deployment. Full article

Vertical farming (VF) offers a smart way to grow crops in stacked layers inside controlled indoor environments. By doing so, it uses far less land and water than traditional open-field agriculture, making it a promising solution for cities with limited space and resources. In recent years, artificial intelligence (AI), machine learning (ML), and Internet of Things (IoT) technologies have begun to transform vertical farming. These tools are moving the industry away from rigid, rule-based systems toward more flexible, data-driven operations that can adapt in real time. This paper presents a systematic review of 208 peer-reviewed studies from 2015 to 2025. It explores how AI, ML, and IoT are applied across the VF ecosystem, focusing on key areas such as computer vision for disease detection, crop growth and yield prediction, smart climate control, and precision nutrient and irrigation management. This review examines the performance of different algorithms, including Convolutional Neural Networks (CNNs), Random Forest, XGBoost, and LSTMs across hydroponic, aeroponic, and aquaponic systems. The review also covers IoT setups with multi-sensor networks, edge-cloud computing, and automated control systems. Commercial farms have shown real gains in resource efficiency and shorter supply chains. However, challenges remain: high energy use (especially from LED lighting, which makes up 40–60% of costs), expensive setup, scattered datasets, and limited real-world testing. Many high-accuracy claims (>95%) come from lab conditions and need better validation in actual farms. Overall, AI-powered vertical farming has strong potential to support resilient urban food systems. Future work should focus on lightweight edge AI models, improved data standards, explainable AI, and robust life cycle assessments to ensure the benefits outweigh the environmental and economic costs. Full article

Federated learning (FL) enables privacy-preserving collaborative training at the network edge, a core capability envisioned for sixth-generation (6G) wireless systems. While surveys and scale-oriented benchmarks advance FL methodology, documented, head-to-head comparisons of mainstream Python frameworks under identical FedAvg settings remain scarce. We benchmark six frameworks—Flower, TensorFlow Federated (TFF), FedML, NVIDIA FLARE, OpenFL, and PySyft—distinguishing two native SDK integrations (Flower, TFF) from four runs of a shared PyTorch FedAvg reference harness (FedML, NVIDIA FLARE, OpenFL, PySyft) in a controlled two-phase study on a Proxmox virtualized testbed with containerized runners, formalize the FedAvg objective and communication-cost model, and position our contribution against prior surveys, scale benchmarks, and single-framework documentation. Each framework–dataset pair is repeated over five IID partitions (random seeds 42–46); we report round-10 mean ± standard deviation for accuracy, wall time, and simulated communication volume. Phase 1 (MNIST) confirms protocol fairness ($99.22\pm 0.07$–$99.29\pm 0.06$% accuracy) with moderate wall-time spread; Phase 2 (CIFAR-10) exposes stack-dependent accuracy gaps (TFF $71.16\pm 0.23$% vs. ≈68% for PyTorch runners). We report per-round accuracy and loss curves with variability bands, wall-time comparisons, and simulated parameter traffic for all six frameworks across nine figures. The experimental protocol, model topology, and hyperparameters are specified in full; per-round JSON metrics and global model checkpoints are published. The study provides a documented baseline for 6G edge framework selection and for follow-on network-constrained and security experiments. Full article

(1) Background: The disaggregated and software-defined nature of fifth-generation (5G) core networks and the Open Radio Access Network (O-RAN) architecture increase the attack surface and produce large volumes of heterogeneous evidence that must be analyzed in real time to support incident reconstruction. Open-source 5G stacks (including Open5GS and srsRAN) have become reference platforms in the literature, yet recent research, such as the RANsacked study that reported 119 vulnerabilities and 97 unique CVEs across multiple LTE/5G implementations, have highlighted the pressing need for AI-based detection and forensic capabilities specific to these stacks. (2) Methods: We introduce an experimental framework consisting of a reproducible srsRAN+Open5GS testbed and an AI-driven forensic and detection pipeline. The pipeline receives control-plane (NAS, NGAP, F1AP) and Service-Based Interface (SBI) traffic, extracts protocol- and statistically grounded features and classifies traffic into seven attack types using a hybrid CNN–LSTM model. Integrity-protected and timeline-correlated forensic artifacts (PCAP, logs, memory dumps) assist in reconstructing an incident. (3) Results: The proposed hybrid model achieves a macro F1-score of 0.972 and an AUC-ROC of 0.995 (5-fold CV) and degrades gracefully under load. (4) Conclusions: We show that AI-based detection can be coupled with a scientifically sound evidence chain in open-source 5G stacks deployed as disaggregated mobile networks. Full article

Robotics simulation is a key enabler for automation and control development, allowing safer experimentation, faster design iterations, and reduced development cost. Nevertheless, the current simulator ecosystem is highly fragmented, spanning open-source and commercial tools with different levels of physical fidelity, performance, and ecosystem integration. As a result, simulator selection is frequently driven by familiarity or availability rather than by explicit task requirements, often leading to suboptimal engineering workflows. This paper proposes a task-oriented decision framework to support reproducible and transparent selection of robotics simulators based on a fixed and structured set of evaluation criteria. These criteria cover (i) physical fidelity and contact modelling; (ii) sensor modelling and visual realism; (iii) performance and scalability aspects, including headless execution, parallelism, and GPU acceleration; (iv) ecosystem integration with automation, control, and learning pipelines, including ROS/ROS 2 compatibility; (v) extensibility and programmability; and (vi) practical constraints such as hardware requirements, licensing models, and learning curve. The framework is operationalised through a checklist and scoring matrix guided by four key questions addressing the target task, fidelity-versus-speed priorities, target software stack, and sim-to-real transfer requirements. To examine feasibility in a representative engineering workflow, a URDF-based modelling and simulation pipeline is implemented and used to compare Gazebo, as an open-source physics-based simulator, against MATLAB/Simulink, representing a commercial model-based simulation environment. The comparison reports practical indicators including setup effort, integration complexity, computational requirements, and runtime behaviour for repeated executions of a representative motion-oriented sequence. The results highlight consistent trade-offs across different user profiles and application needs while also revealing open gaps in the field, notably the lack of unified multi-task benchmarks and joint metrics capable of simultaneously capturing simulation fidelity, computational performance, and sim-to-real transfer effectiveness. Full article

The operational satellite monitoring of pasture biomass requires models that transfer beyond the properties on which they were calibrated. We present a hierarchical, open-source cascade that scales in situ clip-and-weigh biomass (n = 1120 samples across eleven sites on five Queensland properties) through UAV digital-surface-model imagery to Sentinel-2 predictions, using TabPFN—a pre-trained transformer foundation model for small tabular data—as the regressor at all three nested spatial scales. Under a leave-one-site-out (LOSO) protocol on twenty site–date aggregates across nine sites, spectral-only Sentinel-2 models failed to transfer (best $R^2=-0.15$, RMSE 4.62 t ha⁻¹). Appending open climate (Open-Meteo ERA5) and topsoil (SoilGrids 2.0) covariates and evaluating five learners (GBM, RF, XGBoost, TabPFN, and GBM + TabPFN stack) on log-transformed biomass increased LOSO $R^2$ to $-0.05$ and reduced RMSE to 4.43 t ha⁻¹; a leaf-nitrogen growth trajectory predicted by the TabPFN nitrogen regressor from our earlier pasture-chemistry work reduced pixel-level LOSO RMSE by a further 4%. Three alternative covariate classes—BARRA-R2 reanalysis climate, three independent fractional-cover products, and Sentinel-1 C-band SAR backscatter—were tested and rejected, all hitting the same RMSE floor. The symmetric negative results indicate that the residual LOSO ceiling on the current nine-property footprint is a sample-size and optical-saturation limit rather than a feature-engineering one; the most tractable operational path forward is to stratify the production model by climatic zone and Queensland Land Type rather than pursuing further covariates within a single global learner. Full article

Piezoelectric flexure micro-positioning stages are essential micromotion actuators for micro-assembly, atomic force microscopy and nano-manufacturing, but intrinsic hysteresis nonlinearity of piezoelectric stacks distorts the linear voltage-to-displacement mapping and induces significant micro-positioning errors. Conventional hysteresis compensation based on offline-calibrated Krasnosel’skii-Pokrovskii (KP) models cannot adapt to time-varying excitation, whereas state-of-the-art adaptive KP control requires auxiliary dynamic equations and imposes high computational overhead on miniature real-time controllers. To address these limitations, this paper develops a single-degree-of-freedom micromotion positioning device equipped with symmetric two-stage displacement amplification mechanisms and straight circular flexure hinges. ANSYS finite element simulations validate the mechanical stiffness, structural safety and linear amplification characteristic of the micro-positioning stage, achieving a maximum output stroke of 95.95 μm. A discretized KP hysteresis model is constructed to accurately capture the asymmetric rate-dependent hysteresis of piezoelectric stacks. On this basis, a lightweight adaptive inverse control framework is proposed, which realizes online tuning of KP weights through gradient descent iteration only relying on real-time position feedback, eliminating static pre-calibration and extra dynamic correction links. Tracking experiments under 0.1–2 Hz sinusoidal waveforms and 3–7 V variable-amplitude sinusoidal waveforms are implemented. Experimental results show that the proposed approach reduces the root-mean-square error (RMSE) by 7.41–85.65% and the mean absolute percentage error (MAPE) by 7.56–87.81% compared with uncompensated open-loop micromotion control. The combined micro-flexure mechanical design and adaptive hysteresis compensation strategy greatly improves positioning accuracy and anti-interference capacity, offering a low-computation technical route for high-performance micro-positioning systems. Full article

Enterprises increasingly run GPU-bound workloads such as Large Language Model (LLM) inference inside virtualized infrastructure, yet practitioners have little systematic guidance on how the choice of virtualization platform affects GPU passthrough in configuration effort as much as performance. This study’s primary contribution is a structured, paired comparison of configuration complexity and performance across six platforms (Proxmox VM and LXC, native KVM, OpenStack VM and Zun, and Podman) on a single NVIDIA RTX 4500 Ada GPU, emphasizing the under-documented container paths (OpenStack Zun and Podman with the Container Device Interface). The platforms differ most in configuration complexity, where OpenStack, especially Zun, proved the most demanding and Proxmox and LXC the most straightforward. As a confirmatory baseline, for a small-model, low-concurrency vLLM workload that does not saturate the device, exclusive passthrough yields statistically equivalent throughput, duration, and inter-token latency across all platforms, within 2% of bare metal; this equivalence is bounded to the non-saturating regime tested and is not a general claim. Running two containers concurrently (each capped at 45% of GPU memory, compute time-sliced) doubled the offered load and raised aggregate throughput by 12–14%, consistent with idle GPU capacity but not isolating a GPU-sharing benefit. Full article

This article presents a real-time edge-to-mobile melanoma image-analysis prototype combining lesion localization, sensitivity-oriented classification, ensemble fusion, and TensorRT-optimized embedded inference. Six classification architectures were evaluated across HAM10000, ISIC 2019, and ISIC 2020 using a unified training and sensitivity fine-tuning protocol. The selected ResNet-50 and MedFusionNet stack achieved AUC values of 0.985, 0.988, and 0.940, respectively, under the benchmark metadata conditions. At the frozen global threshold of $\theta =0.20$, the corresponding sensitivities were 1.000, 0.981, and 0.948. These results are classification-stage metrics computed on full, uncropped test images; sharpness rejection and YOLO-based localization were evaluated separately. Replacing patient-specific metadata with the neutral deployment vector changed sensitivity by at most one melanoma case per benchmark dataset. On an NVIDIA Jetson Orin Nano, the complete pipeline with TensorRT FP16-optimized classifiers operated at more than 22 FPS with an end-to-end latency of 43.6 $\mathrm{ms}$. The system has not been clinically validated, and the open-source implementation is provided to support reproducibility and further research. Full article