Search Results (259)

Post-quantum cryptography changes the size, timing, and algorithmic context of Transport Layer Security (TLS) handshakes, creating a dynamic normal class for anomaly detectors. This study evaluates an assurance framework, rather than proposing a new classifier, on CIC-PQC_OAV v1: 40,010 sessions represented by 32 encrypted-metadata features. An audit finds 185 exact fingerprints shared across the supplied partitions, affecting 1594 sessions, plus one exact conflicting-label group. We compare the fixed split with five-seed, size-matched sample-stratified, exact-disjoint, raw round-3-disjoint, and IQR-normalized round-3-disjoint protocols, each separating fitting, probability calibration, threshold selection, conformal calibration, and testing. The fixed-split LightGBM $F_1$ is 0.9013; controlled five-seed means cluster at 0.8872–0.8898, showing that its gap is not attributable solely to duplicate control. Condition-disjoint tests reveal heterogeneous anomaly transfer and false-positive rates of 0.9998, 0.9897, and 0.6107 for three unseen valid families. Isotonic calibration yields a Brier score of $0.0221\pm 0.0013$, while five-seed perturbations confirm sensitivity to timing masking and byte scaling. The results support layered, dataset-bounded evaluation combining fingerprint independence, condition holdouts, calibration, selective review, robustness, and explanation. Full article

Equitable access to health-promoting public services is an important component of socially sustainable urban development, yet fitness-facility accessibility is often assessed under homogeneous demand assumptions. This study develops an age-differentiated framework for central Harbin, China, integrating 100 m age-stratified population data, facility functional area, pedestrian-network travel costs, and physical-activity participation rates within an MGH-3SFCA model. Accessibility was evaluated for four age groups at 15 min, with shorter-threshold comparisons, population exposure analysis, bivariate LISA, and sensitivity tests. Adults and students generally formed the relatively higher-accessibility pair, whereas children and seniors remained more disadvantaged. More than 91% of each age group was located in zero, very low, or low accessibility classes. High-intensity ball-sports facilities were dominated by geographic non-coverage, while resistance-training facilities were more widely reachable but mostly at very low levels. The adult–senior zero-accessibility population gap widened from 3.52 percentage points at 15 min to 13.66 at 5 min. High–Low clusters occupied only 5.90–7.39% of land grids but contained 11.77–13.25% of corresponding populations. Broad spatial patterns were stable to alternative composite weighting, while some age contrasts were specification-dependent. The framework supports age-responsive and socially sustainable neighbourhood planning by distinguishing spatial coverage, population exposure, and potential local mismatch. Full article

Background/Objectives: Schatzker type is assigned early, usually from the anteroposterior (AP) radiograph. A single benchmark accuracy cannot say whether a model read the fracture, the anatomy around it, the annotation, or how the archive was assembled. We ran four inexpensive controls to separate those contributions. Methods: We benchmarked a ResNet-50 on PlaTiF, a 2026 public release built for artificial-intelligence research that pairs 421 AP knee radiographs from 186 patients with expert Schatzker labels and per-image tibial segmentations. Evaluation used stratified group five-fold cross-validation grouped by patient, five seeds and balanced accuracy. Inputs were cropped to the expert tibial segmentation shipped with the dataset, an oracle localisation unavailable at deployment. Four controls ran on identical folds: a regression given no pixel content; ablation of the tibial pixels with its complement; a regression on the expert mask alone; and an augmentation audit for label-erasing invariances. Results: Among the 128 fracture patients the network reached 0.345 ± 0.030 six-class balanced accuracy, +0.168 over a non-anatomical baseline fitted on the same folds and the same labels (95% CI +0.106 to +0.230, p = 0.002). Recall was graded: 0.72 for Schatzker VI, 0.11 for V and 0.04 for IV, the last two below chance (0.167). Erasing the tibial pixels left 0.257 ± 0.025, read on its own as the target bone being unused; its complement, the tibia with everything else removed, reached 0.367 ± 0.012, and the whole radiograph, which carries both, only 0.297 ± 0.032 (+0.071 for the tibia alone, 95% CI +0.031 to +0.111, p = 0.008). A regression on the expert mask alone reached 0.213 ± 0.034 and was not distinguishable from the erased model. On fracture versus no classifiable fracture the network reached 0.833 ± 0.028 against 0.814 ± 0.016 for a model given no pixels (p = 0.264), and a coronal computed tomography section accompanied 126 of 128 fracture patients but 24 of 58 others (p = 2.9 × 10⁻¹⁹). Conclusions: Each headline number admitted an explanation other than the fracture in the target bone. An ablation reported without its complement misstated where the signal lay, and an augmentation audit overturned our own explanation for the failure of type IV. Controls of this kind cost minutes, and this study illustrates why they can be informative when a benchmark is built on a retrospective clinical archive. Full article

Network intrusion detectors must train at scale and be retrained as traffic drifts, which stresses pipeline reproducibility and distributed execution. We validate a batch training and retraining pipeline with distributed model fitting, a step towards end-to-end machine-learning operations (MLOps) rather than a complete MLOps implementation. It uses gradient-boosted decision trees with XGBoost-on-Spark on a bare-metal high-performance computing cluster; serving, a model registry, and drift monitoring are outside its scope. We evaluate it on two benchmarks, NSL-KDD and CIC-IDS2017, mapped to five traffic superclasses, along four axes: detection quality, class rebalancing, strong scaling from 1 to 16 nodes, and retraining time. Under a connection-grouped split, Macro-F1 reached 0.90 on CIC-IDS2017 (the more conservative of the evaluated splits), compared with 0.95 under a record-level split, while the ordering of rebalancing paradigms differed by dataset. On the ×512 replica (110 GB), model fitting achieved a 10.64-fold speedup at 16 nodes (95% CI 10.50–10.77), and parallel efficiency increased with data volume. Replacing 25 Gb/s Ethernet with 200 Gb/s InfiniBand over IPoIB left the model-fitting time statistically equivalent within a ±5% margin from 2 to 16 nodes (two one-sided tests over 30 repetitions per point). A complete retraining cycle took under half a per cent of an eight-hour maintenance window. We release the pipeline, curated results, and figure code as a reproducible package. Full article

Background and Aims: Hepatocellular carcinoma (HCC) is frequently diagnosed at an advanced stage, and serum alpha-fetoprotein (AFP), the most widely used biomarker for HCC surveillance, exhibits substantial interpatient variability. However, longitudinal AFP patterns preceding HCC diagnosis and their clinical correlates remain incompletely characterized. We aimed to characterize pre-diagnostic AFP trajectories and identify clinical features associated with rising versus non-rising AFP patterns. Methods: We analyzed 529 patients with newly diagnosed HCC at Songklanagarind Hospital, each with at least three serum AFP measurements obtained within five years before diagnosis. Log-transformed AFP values were modeled using latent class mixed models with natural cubic splines and linear fixed and random effects. The optimal model was selected based on Bayesian Information Criterion (BIC), entropy, and clinical interpretability. Multinomial logistic regression and Cox proportional hazards models were used to evaluate associations between AFP trajectory class, baseline characteristics, and overall survival. Results: A two-class natural spline model (df = 3) provided the optimal fit (BIC = 9003). The non-rising trajectory (n = 440; 83.2%) showed stable AFP levels throughout follow-up, whereas the rising trajectory (n = 89; 16.8%) demonstrated exponential increase beginning 18–24 months before diagnosis. Compared to the non-rising group, patients in the rising group were significantly more likely to present with tumors >2 cm (78% vs. 62%; p = 0.009). The rising trajectory was associated with shorter overall survival in univariable analysis (hazard ratio [HR] 1.33, 95% confidence interval [CI] 1.02–1.73; p = 0.033), but not after adjustment for clinical and tumor characteristics, including Barcelona Clinic Liver Cancer stage (adjusted HR 1.01, 95% CI 0.77–1.33; p = 0.920). Conclusions: Two distinct AFP trajectories precede HCC diagnosis. Most patients did not demonstrate a substantial rise in AFP, underscoring the need for complementary surveillance biomarkers. Prospective validation in independent cohorts is required before AFP trajectory-based approaches can be incorporated into HCC surveillance practice. These findings are exploratory and hypothesis-generating rather than a validated clinical prediction tool. Full article

Smart rings are emerging as a promising class of sensing devices that enable unobtrusive, always-available interaction with Internet of Things (IoT) ecosystems. These wearable devices support intuitive gesture-based control while minimizing user attention and preserving mobility. However, despite rapid advances in sensing hardware and gesture recognition algorithms, little is known about how users associate ring-based shared gestures with smart-home commands. To fill this gap, this paper presents Oringano, a framework for designing and evaluating a vocabulary of shared ring-based gestures for controlling IoT devices in a smart home. These gestures are original in that members of the same group can share the same gestures for the same actions, as well as gestures customized by each individual. The proposed approach combines (i) a synthesis of contemporary smart ring-based gesture interaction literature, (ii) a user-centered requirements elicitation identifying representative smart-home control actions, (iii) a gesture elicitation study involving $N=30$ participants to derive a vocabulary of shared ring-based gestures for 15 IoT control actions, (iv) an empirical analysis of gesture agreement and usability of Oringano, a smartphone prototype for managing shared ring-based gestures, and (v) a set of implications for designing shared gestures for future smart-ring systems. Experimental results demonstrate high agreement for concrete actions such as selection, navigation, and media control, whereas abstract actions exhibit greater variability, highlighting opportunities for personalized gestural interaction. The shared gestures benefit from a higher average agreement rate (+84%), a slightly lower goodness of fit (−13%), and a longer thinking time (+115%) than normal ring-based gestures. The subjective satisfaction resulting from the usability evaluation of Oringano, based on the elicited vocabulary, is overall positive (4.5/5). These results advance the design of next-generation ring-based systems by bridging user-centered gesture interaction with practical sensing technologies for IoT interaction. Full article

Background: Accurate reproduction of occlusal anatomy is a critical determinant of the clinical performance of posterior composite restorations. The present study compares marginal adaptation in Class I composite restorations completed using a Kool-Dam STAMP, a flowable-composite STAMP, or conventional hand-carving techniques. Methods: Sixty extracted caries-free molars were randomly allocated to three experimental groups (n = 20). Two fixed marginal sites per tooth (mesial and distal) were measured at four measurement time points: immediately after restoration, before thermocycling, and after 10,000, 30,000, and 50,000 cumulative thermal cycles (5 °C/55 °C; 20 s dwell time). The two site-level values were averaged within each tooth for the primary analysis. A linear mixed-effects model included technique, measurement time point, and the technique-by-time-point interaction, with a random intercept for tooth. Results: No significant interaction was observed between restorative technique and measurement time point, indicating that changes in marginal gap over time were not statistically significant among the three techniques (likelihood-ratio χ²(6) = 1.43, p = 0.964). The overall effect of the technique was also not significant (χ²(2) = 1.26, p = 0.532). In contrast, marginal-gap measurements increased considerably across the four measurement time points (χ²(3) = 218.52, p < 0.001), indicating progressive increases in marginal gaps with thermal ageing. Following the Holm adjustment, no pairwise comparisons between techniques at any individual time point achieved statistical significance. Conclusions: In the present in vitro study, no statistically significant differences in marginal-gap values were detected among the Kool-Dam STAMP, flowable-composite STAMP, and conventional hand-carving techniques. The study was not designed to demonstrate equivalence or noninferiority; therefore, the absence of a statistically significant difference should not be interpreted as evidence that the techniques are equivalent. These findings suggest that the STAMP approach may be a useful technique for reproducing occlusal morphology while maintaining marginal fit. However, further clinical studies are needed to confirm its long-term performance in vivo. Full article

(1) Background: Phishing remains a pervasive and economically damaging cyberthreat. The dominant detection paradigm has moved toward deep neural and transformer-based classifiers, a literature that reports high accuracy and that does not, in general, expose a per-decision justification, whereas interpretability and auditability are increasingly required in regulated environments; no comparison against transformer-scale detectors is made in this paper. This work asks how far a fully interpretable detector can close the accuracy gap to an opaque text classifier while preserving per-decision explanations, and what such a detector returns that accuracy alone does not measure. (2) Methods: HIEF, an interpretable evidence-fusion framework, is presented. Each email is represented by eighteen human-readable signals: fourteen structural and linguistic cues and four lexical aggregates derived from a published sparse log-odds lexicon. The signals are fused by three transparent layers, namely an $L_1$-regularized logistic model, a shallow interaction-rule tree, and a calibrated Dempster–Shafer stage that reports belief, disbelief and ignorance masses together with an order-invariant global conflict coefficient derived in closed form. A logistic meta-learner fitted on out-of-fold component scores integrates the three layers. The evidential layer uses a type-aware calibration in which discrete signals are calibrated on their attainable values and continuous signals by isotonic regression. Evaluation uses 38,908 public emails, 38,512 of them after exact-duplicate removal, with near-duplicate control, group-aware partitioning, ten repeated splits, a source-held-out protocol, a two-class cross-source test set, a component ablation and a human audit of 100 messages annotated independently by two evaluators. (3) Results: Under group-aware partitioning, HIEF attains an $F_1$ of 0.855 and the strongest term frequency–inverse document frequency (TF–IDF) baseline 0.954; a compact character n-gram neural reference model, evaluated over the same ten partitions, attains 0.973. The linear layer alone attains 0.872, so the two fusion layers do not improve accuracy over it, and the paired difference of 0.017 excludes zero. Type-aware calibration raises the evidential layer from 0.771 to 0.780 and more than halves its partition-to-partition standard deviation, but does not make it competitive; the weakness, therefore, lies in the fusion formulation rather than in the binning. What the evidential layer does supply is a decomposable account of decision uncertainty: the ignorance mass separates errors from correct decisions, 0.265 against 0.175. The human audit reaches an inter-annotator Cohen’s kappa of 0.950 over the five categories before adjudication, and shows that the permissive corpus label agrees with human phishing judgment at a Cohen’s kappa between 0.18 and 0.21, against 0.70 to 0.77 for the automatic strict rule; the audited block is annotated by two of the authors and its human positives are confined to the advance-fee family, so the audit is a bounded comparison of label assignments and not an independent annotation study. (4) Conclusions: HIEF is positioned as an uncertainty and explanation framework rather than as an accuracy-improving fusion method, since the measured accuracy cost of the fusion layers is not compensated by an accuracy gain. Quantifying how much of the performance reported on these widely used corpora is attributable to template leakage and to label permissiveness is a contribution independent of the detector itself. Cross-source operation has not been demonstrated: specificity falls to 0.041 on an unseen collection, so all evaluation reported here is proof-of-concept and no operational deployment claim is made. The Spanish-language evaluation rests on a small and entirely positive subset and is reported as preliminary. Full article

Patients in forensic mental health services (FMHS) with co-occurring mental disorders (MD), substance use disorders (SUD), and a history of violent crime—often referred to as the “triply troubled”—represent one of the most clinically complex and poorly researched patient populations. Standard diagnostics tend to treat this group as homogeneous, which may impede the development of patient-centred treatment and risk management. This study aimed to identify patient subgroups within this population based on patterns of MDs, substance use history, and aggressive antisocial behaviour (AAB). A latent class analysis (LCA) was conducted using clinical data from 119 patients at a high-security FMHS facility in Sweden. Indicator variables included lifetime diagnoses of SUD and MDs and types of substances used. History of AAB and criminal offending were used as distal outcomes. A four-class solution provided the best model fit. Class 1 (polysubstance, high AAB; 100% SUD probability) and Class 4 (polysubstance, bipolar disorder; 49% SUD probability) exhibited pronounced criminal and AAB history. Class 2 (psychosis and cannabis-limited substance use) and Class 3 (autism and limited substance use) represented less criminologically ladened subgroups. Notably, Classes 1 and 4 shared extensive AAB histories despite differing SUD diagnostic probabilities, highlighting the limitations of diagnosis alone as a basis for risk assessment. The four-class typology offers a clinically meaningful framework for stratifying triply troubled patients in FMHS, aligning with the Risk–Need–Responsivity model and severity-graded treatment recommendations The typology has direct implications for treatment differentiation, relapse prevention, and risk management in forensic psychiatric settings. Full article

Reliable coastal wetland habitat mapping is often constrained by the scarcity and the cost of reliable reference data, especially in data-limited coastal environments. We propose a validation-controlled, label-efficient framework pairing multi-season Sentinel-1 and Sentinel-2 predictors with a CatBoost teacher and a lightweight MLP student. A candidate is pseudo-labeled only when both separately calibrated models agree and exceed class-specific thresholds; accepted labels are class-balanced and down-weighted. The framework was evaluated at the Sidi Moussa–Oualidia wetland complex and Merja Zerga lagoon in Morocco. At Sidi Moussa–Oualidia, 62 configurations were compared through nested polygon-grouped validation and then frozen before a five-seed held-out evaluation. The supervised MLP and Agreement-augmented MLP achieved mean Macro-F1 values of $0.9518\pm 0.0044$ and $0.9509\pm 0.0062$, indicating that augmentation did not materially change the already strong full-data baseline. Under a stricter budget of 30 training and 20 validation observations per class, Agreement yielded a mean Macro-F1 of $0.9092\pm 0.0102$ compared with $0.9023\pm 0.0093$ for the supervised baseline and produced pseudo-labels in all five seeds. A spatial-range sensitivity analysis further showed that both models retained Macro-F1 values of 0.9391 and 0.9403 for test observations located beyond the largest estimated within-class autocorrelation range. At Merja Zerga, the native six-class supervised MLP achieved $0.9456\pm 0.0050$, compared with $0.9401\pm 0.0047$ after Agreement augmentation. Spatially blocked four-class experiments nevertheless showed that 20 to 30 local training labels per class recovered approximately 96–98% of the corresponding full-data performance. The framework therefore supplies an operational criterion for using unlabeled observations: augmentation is adopted only where calibrated filtering yields adequate class coverage, and validation confirms a downstream effect; otherwise the supervised model is retained. For the strict Sidi Moussa–Oualidia reduced-label experiment, the reported development budgets count every site-specific label used for fitting, early stopping, and calibration. The Merja Zerga blocked experiments separately quantify training-label sensitivity while retaining their blocked validation resources. Full article

Serial presepsin measurements were evaluated for their prognostic value compared with conventional laboratory markers in predicting in-hospital mortality among older adults with hip fractures. This prospective observational study included 85 patients aged ≥65 years admitted with acute hip fractures between December 2025 and May 2026. Residual serum remaining after routine clinical laboratory sampling was obtained on admission (Day 1), Day 3, and Day 5; serum presepsin was measured by commercial ELISA, and routine parameters (C-reactive protein [CRP], white blood cell count, lymphocytes, monocytes, platelets, and liver enzymes) were analysed in the hospital laboratory. Renal function was assessed by serial creatinine and estimated glomerular filtration rate (eGFR). Patients were classified as survivors or non-survivors according to in-hospital outcome. Temporal trajectories were modelled with a linear mixed-effects model fitted to log-transformed presepsin, receiver operating characteristic (ROC) analysis assessed predictive performance, and internal validity was examined by bootstrap resampling. Of the 85 patients, 68 survived and 17 died during hospitalization; all deaths occurred between hospital days 5 and 12. Presepsin diverged progressively between groups, reaching significantly higher concentrations in non-survivors by Day 5 (median 207.70 vs. 147.21 ng/L; p < 0.001), with a Day 5 group-by-time interaction ratio of 1.76 (95% CI 1.38–2.25; p < 0.001). Day 5 presepsin showed the highest predictive accuracy (AUC = 0.868, 95% CI 0.774–0.945; optimism-corrected AUC 0.866), significantly outperforming CRP (AUC = 0.606; p = 0.003) and platelet count (AUC = 0.485; p < 0.001). The association persisted after adjustment for age, ASA class, fracture type, and eGFR, and Day 5 presepsin added discrimination to a baseline clinical model (ΔAUC 0.125; p = 0.015). The Youden-derived cutoff of 169.54 ng/L was unstable across bootstrap resamples (95% range 169.5–207.7 ng/L). Serial presepsin measurement, particularly on Day 5, is associated with in-hospital mortality in this population; these exploratory findings require external validation before any clinical application can be considered. Full article

The aim of the study is to compare changes in physical fitness in children aged 7–10 training in two martial arts: Korean International Taekwon-Do Federation (ITF) Taekwon-do and Japanese Kyokushin Karate. A group of 163 children (51.5% boys) aged 7–10 years (median-7.7 years (Q1-7.4, Q3-8.1)) was examined. The study participants were divided into: Training group (T = 55), group TT (n = 19, Taekwon-do), group TK (n = 36, Kyokushin Karate) and the control group (C), consisting of 108 children. The study was multi-stage: Stage I—an initial screening for group selection, height, weight, and segmental body composition, as well as physical fitness using selected EUROFIT tests. In Stage II, children from groups TT and TK completed a 12-week training programme, while group C did not change their physical activity routines. After the program was completed, all groups (TT, TK, C) repeated the tests from Stage I. In the group of children training Taekwon-do, positive changes were observed in the range of fitness tests, including flexibility, speed of hand movements and abdominal strength; the differences are statistically significant. In the group of children training Karate, positive, statistically insignificant changes were observed only in functional strength. Introducing martial arts elements into physical education classes can have a positive impact on improving the physical condition of children aged 7–10. Full article

Recommendation systems in public libraries rely on loan data skewed toward past popularity, making it difficult for unborrowed and newly published titles to reach users. Hence, we propose and evaluate a machine learning-based approach predicting library material popularity using open loan data from South Korean public libraries. Three feature sets were constructed: word2vec-based title embeddings (F₁), borrower demographic features (gender and age group; F₂), and topic features from the Korean Decimal Classification (KDC) main class (F₃). Seven machine learning models were evaluated using title-level grouped cross-validation, and XGBoost was selected as the best-performing model. Using this model, the effect and marginal contribution of the feature sets were examined via pairwise t-tests on title-level grouped cross-validation repeated 30 times. Consequently, the full feature set F outperformed all two-feature-set combinations, and F₃ emerged as a key feature set. The feature sets were consistently ranked F₃ > F₂ > F₁ in both predictive performance and model fitness. A cold-start evaluation confirmed near-identical performance for entirely unseen titles. Thus, library material popularity can be predicted using only publicly available information, suggesting the feasibility of a privacy-preserving approach to informing library material recommendations, relevant to library use, digital inclusion, and social justice research. Full article

Objective: To create and test in a population study a behavior score, using as far as possible, procedures based on the a posteriori approach. Material and Methods: Data from the Italian Rural Areas (IRAs) of the Seven Countries Study of Cardiovascular Diseases (SCS), comprising 1712 middle-aged men enrolled in 1960 and their entry levels of smoking habits, physical fitness and dietary habits, were entered into a Principal Components Analysis, producing an individual overall behavior (factor) score. The end-points were the 62-year follow-up, approaching cohort extinction, mortality for all-cause death and five cause-specific conditions, which formed the dependent variables of a series of Cox models, whose outcome was expressed by hazard ratios for one class increase in the score, and comparing tertile 1 versus tertile 3 of the score, without and with the addition of age, systolic blood pressure, serum cholesterol and high socio-economic status as possible confounders. Results: Fatal coronary heart disease (CHD), heart diseases of uncertain etiology (HDUE), stroke, cancer (CAN), chronic bronchitis (CB) and all-cause mortality (ALL) were predicted in a significant way in the four approaches, with a relevant reduction in death rates expressed by hazard ratios ranging from 0.42 to 0.83. Moreover, the influence of the three basic behavior scores showed that CHD was the most strongly associated with them. Age at death for all-cause mortality, on the other hand, was directly related with the full and the three types of behavior score. Conclusions. A behavior score based on three common lifestyle habits created by an a posteriori approach proved to be significantly associated with various causes of death in a male population group followed-up until extinction. Full article

Open-pit mines contain rapidly changing terrain, discontinuous bench structures, and mixed artificial–natural objects, which complicate automated three-dimensional mapping. This study presents a dual-module workflow for UAV LiDAR point clouds. Module A characterizes local geometry using normal and curvature descriptors, constructs local plane support through RANSAC fitting, and detects candidate bench-line points using an angular-gap criterion, followed by regional grouping and Kalman-filter refinement. Qualitative overlay with the orthophoto showed coherent correspondence with principal platform–slope transitions. Module B segments buildings, roads, and vegetation using a PointNet++ network enhanced by local Transformer self-attention and inverted residual feature transformation. Under a fixed spatial hold-out setting, the network achieved an overall accuracy of 97.6% and a mean intersection over union of 96.4%. It obtained the highest overall accuracy, mean intersection over union, and class-wise intersection over union among the selected baselines, whereas Point Transformer achieved a slightly higher mean class accuracy. The two independently operated modules provide complementary structural and semantic information for open-pit mine mapping. Broader applicability requires reference-based bench-line assessment and evaluation across additional mines and survey periods. Full article