Search Results (33)

The rapid growth of the Internet of Things (IoT) has increased dependency on Transport Layer Security (TLS) for securing device communications, enhancing confidentiality while reducing the visibility required by traditional intrusion detection systems. As payload inspection becomes impractical in encrypted environments, anomaly detection must instead rely on flow-level statistics and TLS metadata. This is challenging because IoT traffic is heterogeneous, non-stationary, and distributionally inconsistent across datasets, while many existing studies rely on single-dataset evaluation and therefore provide limited evidence of real-world generalisation. We introduce a TLS-aware anomaly detection framework that combines a $\beta$-Variational Autoencoder ($\beta$-VAE) with a hybrid ANOVA–Mutual Information (ANOVA–MI) feature-selection pipeline. The incremental contribution lies not in the individual use of these components, but in their integrated application to encrypted IoT anomaly detection under strict cross-dataset evaluation, where feature filtering, probabilistic latent regularisation, and threshold transferability are jointly examined without retraining or recalibration on target datasets. The framework models benign encrypted IoT traffic using probabilistic latent representations and identifies anomalies through reconstruction-error-based scoring. Network flows from the BoT-IoT, IoT-23, and ToN-IoT datasets were processed using Zeek and CICFlowMeter to construct a unified metadata feature space incorporating flow statistics and TLS attributes such as JA3 and JA3S fingerprints. The model was trained on benign BoT-IoT traffic and evaluated in both in-dataset and cross-dataset scenarios. The model achieves strong in-dataset performance on BoT-IoT (ROC-AUC $\approx 0.9996$; F1 $\approx 0.9922$) and retains robust anomaly-ranking and threshold-based detection capability under cross-dataset domain shift (IoT-23: ROC-AUC $\approx 0.9882$, F1 $\approx 0.9422$; ToN-IoT: ROC-AUC $\approx 0.9465$, F1 $\approx 0.8732$). A comparative evaluation against deterministic autoencoders and classical baselines further indicates that the proposed $\beta$-VAE achieves stronger cross-dataset anomaly-ranking performance than the compared methods. These findings support the suitability of probabilistic latent modelling for privacy-preserving anomaly detection in encrypted IoT environments. Full article

The automated detection of deep-sea polymetallic nodules is critical for processing large volumes of benthic imagery. However, its scalability faces challenges from cross-expedition covariate shifts, such as changes in lighting, altitude, and camera payloads, which lower zero-shot model performance. While semi-supervised pseudo-labelling presents a potential alternative to time-consuming re-annotation, simple implementations can quickly lead to confirmation bias. This study identifies two primary sources of this degradation: spatial noise from tiling fragmentation at tile borders and an architecture-agnostic interior false positive floor caused by semantic domain shift. This work proposes using a multi-model ensemble for pseudo-labelling to reduce the noise impact. Using a spatial border filter and confidence stratification, three architecturally distinct teacher models (YOLOv8, Faster R-CNN, and DINO) are employed to determine a reliable and domain-invariant subspace. Under a strict anti-leakage Leave-One-Partition-Out protocol, the proposed approach surpasses the supervised fine-tuning baseline at 100-tile pseudo-label budget across four random seeds (macro mAP_50:95 of $0.4745\pm 0.0042$ versus $0.4467\pm 0.0079$), with gains concentrated in the most domain-shifted fold. Beyond this budget, our findings highlight two important adaptation trends: a pool-size degradation trend where excessive pseudo-label volume actively degrades generalisation, and the observation that the fine-tuned models reduce pseudo-label fidelity despite higher precision, providing evidence for the advantage of using frozen source checkpoints for cross-domain adaptation. Full article

Diabetic retinopathy is a leading cause of vision loss worldwide, particularly impacting individuals in low- and middle-income countries with limited healthcare access. Early detection through automated screening systems is essential for improving outcomes, as timely intervention can prevent severe vision impairment. However, most of the available AI models have not been evaluated in low-resource settings. Hence, this study presents an evaluation of the efficacy of advanced deep learning architectures for detecting rDR across diverse population datasets. A dual-phase validation approach was employed to assess model performance. Internal validation utilised the BrSET dataset to establish baseline performance metrics, while external validation was conducted on the MoDRIA dataset, which encompasses various conditions and demographics, to evaluate model robustness. Key performance metrics, including accuracy, specificity, sensitivity, F1-score, and calibration scores, were systematically recorded and analysed. Internal validation revealed high accuracy across all models, EfficientNetB0 achieved the highest classification accuracy (0.9561; 95% CI 0.9490–0.9630), EfficientNetB3 demonstrated superior overall discriminative performance, achieving the highest AUROC (0.9892; 95% CI 0.9841–0.9934) highest sensitivity (0.9573), and lowest Brier score (0.0168). Meanwhile, DenseNet exhibited the most balanced clinical screening performance, achieving the highest F1-score (0.7259; 95% CI 0.6797–0.7669) and Youden Index (0.2381), indicating improved balance between sensitivity and specificity. In contrast, external validation revealed substantial deterioration in model performance across all architectures, highlighting major limitations in cross-population generalisability. Although EfficientNetB0 achieved the highest external accuracy (0.8821; 95% CI 0.8746–0.8898), AUROC values declined markedly across models (0.5140–0.6104), accompanied by poor sensitivity, reduced F1-scores, and substantial calibration instability. EfficientNetB3 achieved the highest external sensitivity (0.5939), whereas calibration analyses demonstrated unreliable probability estimation under domain-shift conditions. These findings suggest that AI models trained on geographically homogeneous retinal imaging datasets may not generalise reliably across underrepresented populations. Population differences and imaging variability substantially affected external model performance, highlighting the need for diverse datasets, rigorous external validation, and adaptive recalibration before clinical deployment of AI-driven DR screening systems. Full article

Background: Celiac disease (CeD) requires lifelong adherence to a strict gluten-free (GF) diet. During the COVID-19 pandemic, the prevailing clinical assumption was that food supply disruptions and dietary management difficulties would be the primary sources of patient distress. This exploratory cross-sectional study directly tested this assumption in an Italian CeD cohort. Methods: COVIMPACT is an exploratory observational, web-based study conducted in Italy (data collected: July–September 2024; participants retrospectively reported their experiences during the COVID-19 pandemic period 2020–2022). Participants with a confirmed CeD diagnosis were recruited through patient associations and online networks. A structured 26-item questionnaire addressed socio-demographic, nutritional, psychological, and healthcare-access domains. Descriptive statistics, chi-square bivariate analyses (Cramér’s V as effect size), and binary logistic regression were performed using R (v4.1) and Python. Results: Among 118 participants (78% female; median age 36 years; IQR 12–42), 27% reported self-perceived psychological distress. Against expectation, difficulties in accessing GF products and changes in gluten consumption showed no clear associations with distress. Instead, social exclusion showed the strongest association (Firth OR = 5.55, 95% CI: 1.80–17.09, p = 0.003), while reduced physical activity (Firth OR = 5.28, 95% CI: 1.86–14.99, p = 0.002, full model; Firth OR = 5.54, p = 0.001, reduced model) and negative economic impact (Firth OR = 3.77, 95% CI: 0.89–15.97, p = 0.071, trend) were additional associated factors. Female sex showed a non-significant trend (Firth OR = 4.21, p = 0.082). All estimates carry wide confidence intervals (EPV = 4.1) and should be treated as hypothesis-generating. Conclusions: These preliminary findings suggest that social exclusion and physical inactivity may be more strongly associated with self-perceived distress than dietary challenges in contexts where GF food access is structurally protected. Results are exploratory, hypothesis-generating, and should not be generalised beyond this selected Italian cohort. Full article

Aircraft target detection in optical remote sensing imagery is hindered by severe scale variation, cluttered backgrounds, and the limited capacity of the spatial-domain convolution to represent frequency-selective target features. We propose FAMA-DET, a frequency-domain adaptive detection framework built on YOLO11, which pursues a unified design principle of content-adaptive spectral representation across all architectural levels. The Frequency-Domain Adaptive Cross-Stage Feature Extractor (FDACFE) replaces static kernels with frequency-domain parameterised convolution driven by learnable DFT basis vectors, enabling differentiated perception of high-frequency edge details and low-frequency semantic components. The Soft-Aligned Bidirectional Feature Pyramid Network (SABFPN) eliminates upsampling amplitude distortion through scale-normalised interpolation and enriches cross-scale fusion with multi-receptive-field textural modelling. The Adaptive Multi-Scale Recalibrated Decoupled Detection Head (AMRDDHead) embeds multi-scale channel recalibration into both localisation and classification branches to suppress background redundancy and reinforce discriminative activations. On MAR20, FAMA-DET improves mAP50 and mAP50-95 over the YOLO11n baseline by 1.8% and 1.6% at only 5.4 GFLOPs, while maintaining real-time throughput of 109.7 FPS. Under zero-shot cross-domain transfer to CORS-ADD, FAMA-DET achieves the highest mAP50 of 93.3% among all compared methods, surpassing RT-DETR-R18 in mAP50 while using 91.0% fewer GFLOPs, confirming that frequency-domain adaptive design yields both strong generalisation and deployment efficiency. Full article

This study investigated the application of six machine learning regression algorithms such as Random Forest, CatBoost, K-Nearest Neighbours, XGBoost, LightGBM, and Gradient Boosting, paired with Molecular ACCess System (MACCS) key fingerprints for the quantitative prediction of aromatase (CYP19A1) inhibitory potency, expressed as pIC₅₀. A dataset of 187 compounds was assembled from the ChEMBL database (version 33, Target ID: CHEMBL1978) following by systematic data curation workflow encompassing duplicate removal, pIC₅₀ transformation, and activity-based filtering. Model performance was rigorously evaluated using an 80/20 stratified train/test split, 5-fold cross-validation, and Y-randomisation testing to ensure unbiased assessment of predictive generalisation. Feature selection via CatBoost permutation importance on the held-out test set identified the top 20 predictive MACCS keys from an initial 166-bit space, substantially reducing dimensionality and improving generalisation across all models. Among the algorithms evaluated, CatBoost trained on the top 20 features achieved the strongest test-set performance (R² = 0.693, RMSE = 0.794, MAE = 0.659) with the most stable cross-validation R² (0.062 ± 0.304), outperforming all other algorithms. Y-randomisation testing returned an empirical p-value of <0.01, confirming that model performance reflects genuine structure–activity relationships rather than statistical chance. Permutation importance and SHAP analysis identified nitrogen-containing heterocyclic fragments (MACCS_41, MACCS_145) and halide-bearing substructures (MACCS_109) as the primary structural determinants of aromatase inhibitory potency, consistent with established CYP19A1 pharmacophoric requirements. Application of the model to ten representative plasticizers demonstrated that the refined applicability domain (h* = 0.423) accommodated eight of the ten compounds, enabling reliable potency predictions across phthalate esters and bisphenol analogues. These findings establish a transparent and reproducible QSAR framework for first-tier endocrine disruption risk screening of plasticizers and highlight the importance of permutation-based feature selection and applicability domain assessment in QSAR model development. Full article

Mental workload assessment using physiological signals has gained increasing attention for applications in human–computer interaction and occupational monitoring. Among these signals, electrodermal activity (EDA) is widely recognised as a reliable indicator of sympathetic activation associated with cognitive effort. However, most existing machine learning- based approaches are evaluated within a single dataset, limiting their generalisability across different populations and experimental conditions. This study investigates the cross-dataset performance of machine learning models for mental workload detection using EDA features. Two independent datasets were employed, and a cross-dataset evaluation framework was adopted to simulate realistic deployment scenarios under domain shift. Three classifiers (Random Forest, XGBoost, and Support Vector Classifier (SVC)) were evaluated, together with two domain adaptation techniques: Correlation Alignment (CORAL) and Subspace Alignment (SA). The results show that model performance is strongly dependent on the direction of transfer, with a notable performance drop when generalising across datasets. Domain adaptation improved performance in several configurations, particularly for SVC with CORAL, achieving the best overall F1-score (0.815). However, improvements were not consistent across all models and target domains. Overall, this study highlights the challenges of cross-dataset generalisation in EDA-based workload detection and demonstrates the potential, yet limited robustness, of domain adaptation techniques in mitigating distribution shifts. Full article

Cross-domain semantic classification of 3D point clouds remains challenging due to strong domain shifts between heterogeneous sensing modalities. Most existing classification frameworks are domain-specific, limiting their use in integrated land–water mapping applications. This study evaluates the transferability of classical geometric machine learning classifiers between terrestrial and underwater point cloud domains without target-domain retraining. Experiments were conducted using terrestrial data acquired with a Leica BLK360 terrestrial laser scanner (TLS) and underwater point clouds collected with a Blueview BV5000 mechanical scanning sonar (MSS). Two dimensionality-based frameworks, CANUPO–Support Vector Machine (SVM) and 3DMASC–Random Forest (RF), were implemented in CloudCompare and assessed under intra-domain and cross-domain configurations. Strong intra-domain performance was achieved, with terrestrial–terrestrial accuracies of 0.99 for CANUPO–SVM and 0.97 for 3DMASC. In underwater evaluation, CANUPO maintained high accuracy (0.97), whereas 3DMASC decreased to 0.86 due to increased variability in the submerged data. Under cross-domain transfer, CANUPO achieved 0.93 accuracy for terrestrial-to-underwater and 0.89 for underwater-to-terrestrial classification, while 3DMASC demonstrated stable generalisation with 0.95 accuracy in both directions. Overall, dimensionality-based geometric descriptors capture stable structural cues across sensing environments, providing an interpretable and efficient pathway for applications such as hydrographic surveying, coastal monitoring, and underwater search-and-rescue detection. Future work will extend validation to larger datasets and explore domain adaptation strategies to further reduce cross-modality domain shift. Full article

This paper investigates federated multi-centre liver tumour classification from contrast-enhanced CT under realistic data heterogeneity and domain shift. To address the practical constraint that medical data are often siloed across institutions, we develop a FedProx-based federated learning pipeline that enables collaborative training without exchanging raw patient data. Using the LiTS dataset as the training domain, we construct a slice-level binary classification task based on voxel-level annotations, while rigorously assessing out-of-distribution generalisation on an external held-out dataset, 3D-IRCADb. We conduct comprehensive experiments across multiple backbone architectures, including ResNet-50, EfficientNet-B3, ViT-B/16, and MobileNetV3-Small, comparing FedProx and FedAvg under three heterogeneity intensities (IID, mild non-IID, and severe non-IID). Furthermore, we evaluate transfer learning strategies, ranging from frozen backbones to partial fine-tuning of the last stage, and perform ablations on the proximal coefficient $\mu$ and local epochs E to characterise optimisation behaviour. Our results show that FedProx is generally comparable to FedAvg, with slightly more stable behaviour in some heterogeneous settings. We also observe a clear validation-to-external gap, indicating that external-domain robustness remains challenging and requires cautious interpretation for deployment. ImageNet pretraining yields consistent gains, particularly for data-sparse clients, while partial fine-tuning enhances adaptation to CT-specific features. Finally, MobileNetV3-Small offers a favourable performance–efficiency trade-off by reducing communication payload and computation cost, supporting practical deployment on resource-constrained clinical edge devices. Full article

For Unmanned Aerial Vehicles (UAVs), optical sensing for reliable landing and the detection of the landing area is a crucial element. In low-light conditions, at night, and in foggy weather, where optical sensing is not feasible, thermal imaging can be utilised. Although this situation has been widely researched, most UAV landing approaches rely on GNSS assistance or single-mode detection, which limits their robustness and scalability in real-world operations. This study proposes an actively heated thermal helicopter landing pad designed using electrically powered resistive heating elements and a high-emissivity surface coating. Furthermore, optical and thermal images collected during actual UAV flight experiments under daytime and night-time conditions were processed using image fusion techniques with AVGF, DWTF, GPF, LPF, MPF, and HWTF fusions, and their performance in deep learning models was compared. The obtained optical, thermal, and fused datasets are used to train and evaluate deep learning-based helicopter landing pad detection models based on the YOLOv8 architecture. Experimental results show that models trained with single-mode data exhibit limited cross-domain generalisation, while fusion-based learning significantly improves detection robustness in optical and thermal domains. Among the evaluated methods, LPF, MPF and HWTF provide the most consistent performance improvements. The findings indicate that electrically heated thermal helicopter landing pads, when combined with image fusion and deep learning-based detection, can increase the landing detectability of UAVs at night and in low-visibility conditions. This detection-focused approach contributes to UAV flight safety by enhancing the visibility of the landing area without relying on active infrared markers or additional navigation infrastructure. Full article

Net blotch, caused by Pyrenophora teres, is a major barley disease that occurs in two forms, spot form net blotch (SFNB) and net form net blotch (NFNB), reducing grain yield and quality worldwide. Accurate detection is critical for disease management and breeding resistant cultivars; however, traditional disease scoring is labour-intensive and error-prone. This study evaluates the use of UAV-based red–green–blue (RGB) and multispectral imagery, combined with machine learning, for determining net blotch infection levels at the plot scale across multiple sites and seasons in Australia. Various colour features, vegetation indices, and algorithms were tested, including a cross-domain testing for model generalisation. We propose a robust UAV-driven pipeline enabling precise disease monitoring and phenotyping in barley breeding programs. Full article

Self-supervised contrastive learning has become an effective approach for visual representation learning when large-scale annotation is impractical. In this study, we evaluate three widely used methods—SimCLR, MoCo v2, and BYOL—for large-scale stock keeping unit (SKU) recognition in retail environments. Experiments are conducted on the RP2K benchmark and a domain-specific in-house dataset (InSKU) using both linear probing and full fine-tuning. Under the original RP2K configuration with extended self-supervised pre-training, SimCLR achieves the highest Top-1 accuracy under linear evaluation (94.98%). In contrast, BYOL attains the highest performance under full fine-tuning (99.22% Top-1 accuracy). After filtering and deduplicating the dataset to reduce class imbalance and near-duplicate samples, MoCo v2 achieves competitive, and in some cases superior, linear performance under a reduced training budget. Cross-domain evaluation on InSKU indicates that SimCLR generalises more effectively under frozen-encoder constraints, whereas BYOL and MoCo v2 require full adaptation. These results highlight the sensitivity of contrastive representations to dataset composition, optimisation regime, and domain shift, providing practical guidance for deployment in dynamic retail settings. Full article

In photovoltaic remote sensing image segmentation tasks, fully supervised methods can achieve high accuracy. However, the high cost of pixel-level annotation significantly limits their scalability in large-scale scenarios. To overcome this annotation bottleneck, this paper proposes a zero-shot cross-modal segmentation framework based on the visual-language pre-trained foundation model (CLIP). This approach harnesses CLIP’s cross-modal knowledge transfer capabilities to achieve precise extraction of photovoltaic targets without requiring any downstream training. This paper first introduces the Layer-wise Augmented Residual Attention (LARA) mechanism to enhance fine-grained detail representation in the feature space. Subsequently, a Cross-modal Semantic Attribution Module (CMSA) is designed to generate precise activation maps by leveraging image-text alignment gradient information. Finally, the Confidence-Aware Refinement Strategy (CARS) replaces the conventional training-based denoising process, directly producing high-quality binary segmentation masks through adaptive thresholding. Comparative experiments were conducted to evaluate the proposed method against various baselines using several public datasets with varying resolutions in Jiangsu Province including Unmanned Aerial Vehicles imagery, Beijing-2, Gaofen-2, and a self-created Sentinel-2 imagery covering multiple countries. Notably, the proposed method achieved an IoU of 70.3% on the Gaofen-2 PV03 dataset with a spatial resolution of approximately 0.3 m and 50.8% on the self-created Sentinel-2 PV_Sentinel-2 dataset with a spatial resolution of 10 m. Experimental results demonstrate that our proposed approach maintains excellent cross-domain generalisation capabilities while reducing annotation costs, thereby providing an efficient and viable technical pathway for the automated monitoring of large-scale photovoltaic facilities. Full article

Modern Standard Arabic (MSA) and the many regional dialects differ substantially in vocabulary, morphology, and pragmatic usage. Most available annotated resources are in MSA, and zero-shot transfer from MSA to dialectal tasks suffers a large performance drop. This paper addresses generalised cross-dialectal Arabic question answering (QA), where the context and the question are written in different Arabic varieties. We propose a training-free augmentation framework that generates code-mixed questions to bridge lexical gaps across Arabic varieties. The method produces semantically faithful, balanced code-mixed questions through the following two-stage procedure: lexicon-based partial substitution with semantic similarity and substitution-rate constraints, followed by fallback neural machine translation with word-level alignment when needed. We also introduce automated multidialectal lexicon construction using machine translation, embedding-based alignment, and semantic checks. We carry out our evaluation in a zero-shot setting, where the model is fine-tuned only on MSA and then tested on dialectal inputs using ArDQA, covering five Arabic varieties and three domains (SQuAD, Vlogs, and Narratives). Experiments show consistent improvements under context-question dialect mismatch as follows: +1.09 F1/+0.87 EM on SQuAD, +1.54/+1.25 on Vlogs, and +2.75/+2.27 on Narratives, with the largest gains for Maghrebi questions in Narratives (+12.13 F1/+8.45 EM). These results show that our method improves zero-shot cross-dialectal transfer without fine-tuning or retraining. Full article

Remote sensing plays a vital role in monitoring environmental change in Antarctica, offering non-invasive insights into ice dynamics, biodiversity, and fragile ecosystems. Harsh conditions, limited field access, and logistical challenges result in sparse, noisy, and often unlabelled datasets, posing major obstacles for machine learning (ML) approaches. Data scarcity remains a fundamental challenge for uncrewed aerial vehicle (UAV)-based ecological monitoring. While ML models in other Earth observation domains demonstrate state-of-the-art performance, their applicability in Antarctic and polar regions’ settings is limited. This paper reviews the intersection of ML and UAV-based remote sensing in Antarctica under extreme data constraints. We surveyed recent strategies designed to overcome these limitations, including self-supervised learning, physics-informed modelling, and foundation models. Results highlight a notable gap, as polar environments remain excluded from global datasets and benchmarks due to the extensive data requirements of large-scale models. Opportunities exist where multimodal and multi-scale generalisation can enhance cross-domain adaption to data-scarce use cases. Unlike prior reviews on general remote sensing or task-specific polar studies, this work uniquely underscores the need for Antarctic representation in global ML advances, positioning Antarctica as a frontier testbed for machine learning in extreme, inaccessible, and under-resourced fields. Full article