Hung Yen University of Technology and Education

The transition to 6G and Open RAN (O-RAN) necessitates intelligent orchestration across multi-operator networks, yet this collaboration introduces severe security and privacy risks. Malicious operators may poison global models through adaptive attacks, while the exchange of raw gradients threatens data sovereignty. In this paper, we propose FORTRESS-FL , a robust and privacy-preserving federated learning framework designed for secure cross-domain orchestration. At its core is the TrustChain protocol, which synergizes a commit-then-reveal scheme to prevent adaptive manipulation, unsupervised spectral clustering for Byzantine detection, and a dynamic reputation system to isolate malicious actors. Furthermore, we integrate an adaptive Differential Privacy (DP) mechanism to rigorously protect operator data. Extensive evaluation on a real-world financial fraud dataset demonstrates that FORTRESS-FL achieves 100% detection accuracy against sign-flip attacks with 30% Byzantine adversaries, preventing the model divergence observed in standard baselines. Scalability tests confirm linear complexity with respect to the number of operators, validating the framework’s feasibility for large-scale, real-time network orchestration.

Accurate prediction of individual treatment effects in survival analysis is often complicated by the presence of competing risks and the inherent unobservability of counterfactual outcomes. While machine learning models offer improved discriminative power, they typically lack rigorous guarantees for uncertainty quantification, which are essential for safety-critical clinical decision-making. In this paper, we introduce CONFIDE (CONFormal Inference for Distribution-free Estimation), a novel framework that bridges causal inference and conformal prediction to construct valid prediction sets for cause-specific cumulative incidence functions. Unlike traditional confidence intervals for population-level parameters, CONFIDE provides individual-level prediction sets for time-to-event outcomes, which are more clinically actionable for personalized treatment decisions by directly quantifying uncertainty in future patient outcomes rather than uncertainty in population averages. By integrating semi-parametric hazard estimation with targeted bias correction strategies, CONFIDE generates calibrated prediction sets that cover the true potential outcome with a user-specified probability, irrespective of the underlying data distribution. We empirically validate our approach on four diverse medical datasets, demonstrating that CONFIDE achieves competitive discrimination (C-index up to 0.83) while providing robust finite-sample marginal coverage guarantees (e.g., 85.7% coverage on the Bone Marrow Transplant dataset). We note two key limitations: (1) coverage may degrade under heavy censoring (>40%) unless inverse probability of censoring weighted (IPCW) conformal quantiles are used, as demonstrated in our sensitivity analysis; (2) while the method guarantees marginal coverage averaged over the covariate distribution, conditional coverage for specific covariate values is theoretically impossible without structural assumptions, though practical approximations via locally-adaptive calibration can improve conditional performance. Our framework effectively enables trustworthy personalized risk assessment in complex survival settings.

Public WiFi networks deployed across high-density transit environments — such as train stations, bus terminals, and airports — present a largely unaddressed frontier in network intrusion detection research. These venues collectively serve tens of millions of users monthly yet exhibit extreme traffic heterogeneity, continuous attack pattern drift, and strict data privacy constraints that preclude centralized detection paradigms. Existing federated learning-based intrusion detection systems (FL-IDS) fail to simultaneously address the spatiotemporal dynamics of evolving network topologies, the interpretability requirements of operational security teams, the concept drift inherent to live deployments, and the fairness disparities that arise across heterogeneous venue types. We propose ST-FedXIDS, a Spatiotemporal Federated Explainable Intrusion Detection System that unifies five technical contributions: (1) a heterogeneous dynamic graph construction pipeline that models multi-protocol WiFi traffic as sliding temporal-window graphs; (2) ST-GraphSAGE with multi-head temporal self-attention, replacing the LSTM/GRU paradigm used in all prior spatiotemporal GNN-IDS works; (3) Drift-Adaptive Federated Averaging (DA-FedAvg) that embeds ADWIN-based graph embedding drift detection directly into the federated aggregation protocol; (4) a GNN-Explainer module that produces subgraph-level attack attribution explanations for operational incident response; and (5) a venue-type fairness regularizer that minimizes the maximum inter-venue false positive rate gap across heterogeneous deployment sites. Extensive experiments on three public benchmarks — CSE-CIC-IDS2018, TON_IoT, and UNSW-NB15 — against four 2025 state-of-the-art baselines demonstrate that ST-FedXIDS achieves 98.21%–99.13% accuracy, outperforming all competitors by 0.9–2.4 percentage points, reduces the inter-venue FPR gap from 8.5–12.3% to 2.1%, and recovers from concept drift in 8 federation rounds versus 19–25 rounds for competing methods. An ablation study validates the independent contribution of each component.

Despite the growing burden of comorbidities, including cancer, among people with HIV, there is a lack of empirical evidence on cancer-related mortality in this vulnerable population in Vietnam. This study aimed to examine cancer mortality rates among people with HIV in Vietnam by cancer type, age and sex and compare them with those in the general population. We analysed nationally representative mortality data from 2005 to 2021, collected via Vietnam's mandatory mortality registration system. Cancer deaths and HIV status were identified using the ICD-10 codes. People with HIV data were linked to the Vietnam Ministry of Health HIV/AIDS registry. Mortality rates (per 100 000 person-years) and mortality rate ratios (MRRs) with 95% confidence intervals (CIs) were calculated, stratified by age and sex and age-standardized using the World Health Organization standards. Among people with HIV, 253 cancer deaths occurred across 523709 person-years (crude rate 48.3 per 100 000 person-years), compared to 173 592 cancer deaths over 272 million person-years in the general population (crude rate 63.8 per 100 000 person-years) between 2005 and 2021. The age-standardized cancer mortality rate among people with HIV was 50.5 per 100 000 person-years, compared with 110.2 per 100 000 person-years in the general population. However, age-specific analysis revealed that cancer mortality rates among people with HIV under age 50 were significantly higher than those in the general population (MRR = 2.29; 95% CI: 2.01-2.59), predominantly for lung cancer (MRR = 10.5; 95% CI: 8.89, 12.41) and liver cancer (MRR = 1.82; 95% CI: 1.38, 2.40). The highest MRR was observed in the youngest age group (1-29 years; MRR = 5.67; 95% CI: 4.63, 6.93). Although the age-standardized cancer mortality rate among people with HIV was lower than in the general population, people with HIV experience substantially higher cancer mortality among those under age 50, particularly from lung and liver cancers. These findings emphasize the urgent need for integrated cancer prevention, early detection and treatment strategies for people with HIV in Vietnam.

To determine morbidity, mortality, and risk factors of severe disease requiring invasive mechanical ventilation (IMV) in neonates with pneumonia in rural Vietnam. A retrospective study was conducted and included all neonates (<28 days) admitted with pneumonia between January and December 2024. Demographics, clinical features, and laboratory results were extracted from standardized patient charts. Univariable and multivariable logistic regression were used to identify independent risk factors for IMV. Of 1,034 admissions, 612 (59.2%) met pneumonia criteria; mortality rate was 0.2%. Median age at admission was 17 days (IQR 12-23), and 61.6 % were boys. Preterm birth and low birth weight occurred in 4.6% and 6.4% of cases, respectively. Tachypnoea (96.1%), wheeze (93.0%), and signs of respiratory distress (37.7%) were predominant. Abnormal neutrophil counts were observed in 30.2%, and 5.2% had C-reactive protein (CRP) ≥10 mg/L. Independent risk factors for IMV were younger age (adjusted odds ratio [aOR] 0.88 per day; 95% CI 0.85-0.92), neurological symptoms (aOR 24.0; 95% CI 4.07-141.67), and CRP ≥10 mg/L (aOR 3.23; 95% CI 1.22-8.52). Pneumonia remains a major cause of neonatal morbidity in rural Vietnam, though mortality is low. Younger age, neurological compromise, and elevated CRP identify neonates at greatest risk for requiring IMV. Incorporating these factors into admission triage may streamline escalation of care and optimize resource allocation in neonatal units.

ObjectivesTo evaluate the accuracy of the arterial oxygen partial pressure/inspired oxygen fraction (PaO2/FiO2) ratio in predicting mortality among acute respiratory distress syndrome (ARDS) patients in Vietnam.DesignA retrospective observational study.SettingA central hospital in Vietnam.ParticipantsAdult patients diagnosed with ARDS based on the Berlin definition and admitted to Bach Mai Hospital between August 2015 and August 2023. ARDS severity was converted from descriptive categories to the Berlin score, ranging from 1 (PaO2/FiO2>300 mm Hg) to 4 (PaO2/FiO2≤100 mm Hg).Primary outcomeAll-cause hospital mortality.ResultsOf 345 patients, 67.5% were male, and the median age was 55.0 years (IQR: 39.0–66.0). Hospital mortality was 61.2% (211/345). On the first day of admission, the PaO2/FiO2 ratio (areas under the receiver operating characteristic curves (AUROC): 0.585 (95% CI 0.522 to 0.649)) showed limited predictive ability for hospital mortality. Incorporating the PaO2/FiO2 ratio into the Berlin score did not substantially improve accuracy (AUROC: 0.578 (95% CI 0.516 to 0.641)). Both measures were less accurate than Sequential Organ Failure Assessment (SOFA) (AUROC: 0.650 (95% CI 0.590 to 0.711)), Acute Physiology and Chronic Health Evaluation II (APACHE II) (AUROC: 0.685 (95% CI 0.628 to 0.742)) and Confusion, Urea >7 mmol/L (20 mg/dL), Respiratory rate ≥30 breaths/min, Blood pressure (systolic <90 mm Hg or diastolic ≤60 mm Hg) and Age ≥65 years (CURB-65) (AUROC: 0.689 (95% CI 0.617 to 0.762)). Higher PaO2/FiO2 values (adjusted OR, AOR: 0.988 (95% CI 0.979 to 0.996)) were independently associated with lower mortality risk, while higher Berlin (AOR: 2.477 (95% CI 1.190 to 5.156)), SOFA (AOR: 1.278 (95% CI 1.102 to 1.482)), APACHE II (AOR: 1.236 (95% CI 1.108 to 1.379)) and CURB-65 (AOR: 7.142 (95% CI 2.581 to 19.763)) scores were associated with increased mortality risk.ConclusionsIn this study of ARDS patients in Vietnam, the PaO2/FiO2 ratio demonstrated limited discriminatory ability for hospital mortality, and incorporating it into the Berlin score did not meaningfully improve performance. While less accurate than SOFA, APACHE II and CURB-65 scores, the PaO2/FiO2 ratio and Berlin score remained independently associated with mortality risk. These findings should be interpreted cautiously, given the retrospective design, single-centre setting and potential selection bias; further validation in larger, multicentre studies is warranted.