Abdulaziz A. Alsulami

Local label manipulation can threaten federated medical-image classification. We evaluated Pneumonia-to-Normal poisoning in a client-local feature-federated ensemble using 20,795 chest X-rays from three sources. Three of 10 simulated clients with differing class distributions changed all Pneumonia classifier-training labels to Normal. Across 10 seeds, matched attack and zero-poison runs kept perceptually linked images within the same partition. The vulnerable configuration used first-in-first-out retention, equal weighting, and no tuning. The protected configuration used local-validation retention, validation weighting, and class-score tuning. Their mean attack-minus-control increases in Pneumonia-to-Normal error were 12.12 and 0.79 percentage points, respectively, with 95% confidence intervals of 7.51–16.72 and 0.14–1.43. The paired reduction was 11.33 points, with a 95% confidence interval of 6.86–15.80. Without retraining, tuning alone and retention with weighting alone left increases of 0.91 and 4.11 points, respectively. Complete server-validation Pneumonia-to-Normal contamination in protected control and attacked runs produced an increase of 21.41 points when tuning averaged F1 over classes present in validation labels or predictions, versus 17.90 over all four classes. Respective 95% confidence intervals were 13.17–29.65 and 9.53–26.27. This scorer sensitivity limits contamination-effect interpretation. Inference remains exploratory and conditional on the selected attack direction. Simulated clients, unavailable patient identifiers, and lack of source-held-out testing limit generalization.

Standard federated averaging gives clients with larger training datasets greater weight. How reducing these weight differences affects overall and client-level short message service (SMS) spam detection remains unclear. We compare nine federated methods using logistic regression on 5023 ExAIS messages from 20 clients, defined by source files. The methods include equal weighting, weighting proportional to training size, and square-root weighting to reduce differences in aggregation weights. Evaluation repeats two-fold cross-validation five times, assigning identical normalized messages entirely to the training or testing set in each split. With class-weighted loss, three local epochs per round, and a threshold of 0.5, square-root weighting has the highest mean pooled F1 (0.8150), calculated from all test messages together within each fold. However, none of the four primary comparisons is significant after multiple-testing correction. Equal-client weighting has the highest observed average client F1 and mean F1 for the lowest-performing quarter of clients. Thresholds selected on inner validation data reduce the square-root weighting’s advantage over class-weighted proportional aggregation from 0.0112 to 0.0010. With equal local update counts per client and round, proportional weighting ranks first on pooled F1, followed by square-root and equal-client weighting. In this equal-step analysis, only the square-root versus equal-client contrast is significant after correction for the two post hoc comparisons. On an external SMS corpus, equal-client weighting ranks first among the three rules, with a mean pooled F1 of 0.3960. No aggregation rule performs best across all evaluated settings, and rankings depend on local training, classification thresholds, and whether performance is assessed across messages or clients.

Next-day stock direction forecasting combines market signals with news, but many trading days contain no firm-specific articles. We develop Gated News-Event Fusion (GNEF), an availability-aware model mapping price sequences, engineered market features, and news features to direction probabilities and a nonnegative return-magnitude estimate. GNEF retains price as a residual representation and uses a scalar sigmoid gate to interpolate between technical and news representations. We prove that the gated component is a norm-bounded convex combination whose gate-logit sensitivity is at most one quarter of the distance between these representations. We also establish conditional Lipschitz stability and forward complexity linear in sequence length. GNEF has 67,625 trainable parameters. Evaluation covers ten large-cap equities under prospective expanding walk-forward testing, paired moving-block bootstrap inference, and transaction-cost analysis. Across five random initializations, the residual design has higher mean directional accuracy than a gated mixture of price, technical, and news representations on 7 of 10 tickers. In a single-seed ablation, full GNEF exceeds no-gate and no-news variants in directional accuracy on 3 of 10 tickers. A broader validation-selected policy exceeds momentum in directional accuracy on 5 of 10 tickers. These findings provide selective support for residual fusion but do not establish a consistent advantage across equities.

The increasing popularity of online learning platforms has led to vast amounts of sequential data regarding learner behavior. Available predictive models tend to focus on fixed features. They cannot pinpoint the dynamic temporal changes in learning activity, thereby reducing the effectiveness of predictive methods and making them more challenging to understand. In this work, the Dynamic Temporal Attention Network (DTAN) is proposed. This novel deep learning architecture learns e-learning behavior using time-aware attention and temporal convolution to enhance predictive accuracy and interpretability. TCN has already been combined with two significant attention modules. The Attention-over-Time windows (ATW) and the Dynamic Contextual Attention Mechanism (DCAM). With such components, the model can learn short and long-term dependencies on the learner’s behavior and adaptively prioritize critical time slots. To train and evaluate the model, two large-scale datasets are used: EdNet, containing over 130 million question and answer interactions in the K-12 context, and OULAD, an exam-taking dataset in the university context. By outperforming state-of-the-art models, including long short-term memory (LSTM) and gated recurrent units (GRU), as well as standard TCNs, TAN achieves significant improvements across diverse classification tasks. It has strong early prediction skills and an interpretable visualization that focuses on weight, highlighting critical incidents in a growing learner journey. Such observations are essential in the process of providing individual instruction and necessary intervention. DTAN delivers an interpretable solution for sequential data modeling in education, significantly boosting efficiency, particularly in adaptive learning systems.

Healthcare systems worldwide face increasing pressure to deliver accurate, affordable, and scalable diagnostic services while maintaining long-term sustainability. Chest X-ray screening is considered one of the most cost-effective methods for detecting lung disease. However, many deep learning approaches are computationally intensive and difficult to interpret, which limits their adoption in high-throughput, resource-constrained clinical settings. This study proposes a hybrid CNN–tree framework for automated lung disease classification from chest X-ray images, which targets COVID-19, pneumonia, tuberculosis, lung cancer, and normal cases. To ensure robustness and generalization, four publicly available chest X-ray datasets from different sources are merged into a unified five-class dataset, which introduces realistic variations in imaging conditions and patient populations. A ResNet-18 model is fine-tuned to extract domain-specific deep feature representations. Feature dimensionality and redundancy are reduced using Principal Component Analysis, while class imbalance is addressed through the Synthetic Minority Over-sampling Technique. The resulting compact feature vectors are used to train interpretable tree-based classifiers, which include Decision Tree, Random Forest, and XGBoost. Experiments conducted using five-fold stratified cross-validation demonstrate substantial and consistent performance gains. When trained on fine-tuned and preprocessed deep features, all evaluated tree-based classifiers achieve weighted F1-scores between 0.977 and 0.982 using five-fold cross-validation, with a significant reduction in inter-class confusion. In addition, the proposed framework maintains low per-sample inference latency, which supports energy-efficient and scalable deployment. These results indicate that combining deep feature learning with interpretable tree-based models provides a practical and reliable solution for sustainable chest X-ray screening in real-world clinical environments.

The widespread connectivity of the Industrial Internet of Things (IIoT) improves the efficiency and functionality of connected devices. However, it also raises serious concerns about cybersecurity threats. Implementing an effective intrusion detection system (IDS) for IIoT is challenging due to heterogeneous data, high feature dimensionality, class imbalance, and the risk of data leakage during evaluation. This paper presents a leakage-safe hybrid intrusion detection framework that combines text-based and numerical network flow features in an IIoT environment. Each network flow is converted into a short text description and encoded using a frozen Large Language Model (LLM) called the Bidirectional Encoder Representations from Transformers (BERT) model to obtain fixed semantic embeddings, while numerical traffic features are standardized in parallel. To improve class separation, class prototypes are computed in Principal Component Analysis (PCA) space, and cosine similarity scores for these prototypes are added to the feature set. Class imbalance is handled only in the training data using the Synthetic Minority Over-sampling Technique (SMOTE). A Random Forest (RF) is used to select the top features, followed by a Histogram-based Gradient Boosting (HGB) classifier for final prediction. The proposed framework is evaluated on the Edge-IIoTset and ToN_IoT datasets and achieves promising results. Empirically, the framework attains 98.19% accuracy on Edge-IIoTset and 99.15% accuracy on ToN_IoT, indicating robust, leakage-safe performance.

Machine learning-based intrusion detection for Internet of Things (IoT) networks remains difficult because modern traffic is highly imbalanced and attack behaviors are heterogeneous. Evaluation pipelines can also overestimate performance when preprocessing is performed before train-test separation. We propose a family-aware hierarchical intrusion detection framework for attack-family prediction. The proposed approach first separates normal and attack traffic, then routes attack samples into empirically defined majority and minority attack-family branches, and finally performs branch-specific family classification. Within each cross-validation fold, training-label counts define the majority/minority routing branches, while scaling, weighting, model fitting, stage diagnostics, and metric computation remain fold-local. The final implementation uses XGBoost as the base learner in the hierarchical stages and compares it with flat LightGBM, XGBoost, Random Forest, Extra Trees, and stacking baselines under matched folds and metrics. On the CICIoT2023 30% stratified development split, the proposed approach achieved a Macro-F1 of 0.8380 and Weighted-F1 of 0.9940, performing close to the best flat Random Forest baseline while improving weak rare-family F1 scores for BruteForce and Web. On Edge-IIoTset, where the full processed dataset is used, the proposed approach achieved a Macro-F1 of 0.9456 and Weighted-F1 of 0.9494, outperforming all individual flat baselines and approaching the flat stacking ensemble. The hierarchy had lower inference time than the evaluated flat baselines under the workstation protocol.

Intrusion Detection Systems (IDS) play a critical role in protecting networked environments from cyberattacks. They have become increasingly important in smart environments such as the Internet of Things (IoT) systems. However, IDS for IoT networks face critical challenges due to hardware constraints, including limited computational resources and storage capacity, which lead to high feature dimensionality, prediction uncertainty, and increased processing cost. These factors make many conventional detection approaches unsuitable for real-time IoT deployment. To address these challenges, this paper proposes an adaptive intrusion detection framework that intelligently balances detection accuracy and computational efficiency. The proposed framework integrates mutual information (MI) feature selection model, deep contextual embeddings, and an adaptive decision mechanism. The MI model identifies and retains the most informative features, which reduces dimensionality while maintaining high detection accuracy. The adaptive decision dynamically selects between multiple inference paths to ensure that additional computation is needed only when the uncertainty level is high. Experimental evaluations on benchmark IoT datasets namely RT-IoT-2022, CIC-IoT-2023 and CIC-IoMT-2024 show that the proposed framework achieves F1-score of 99.92%, 96.66%, and 99.84%, respectively, with an average inference time of approximately 0.105 ms per sample. These results demonstrate that the framework effectively adapts inference complexity to data uncertainty, which provides an intelligent, interpretable and efficient solution for real-world IoT intrusion detection.