Computers doi: 10.3390/computers15060354

Authors: Yousef Sanjalawe Budoor Allehyani Sharif Naser Makhadmeh Salam Al-E’mari Ola Surakhi Dima Suleiman

The growing fragmentation of digital evidence in modern computing environments poses significant challenges for digital forensic analysis. Data is often deleted, overwritten, or distributed across heterogeneous platforms, limiting the effectiveness of traditional forensic tools that rely on intact files and deterministic rules. This work addresses a key limitation in current forensic methodologies: the scarcity of learning-based approaches capable of identifying patterns in fragmented and incomplete digital evidence. To address this challenge, we propose PatternMiner, a hybrid deep learning framework that integrates Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Transformer encoders. The framework combines byte-level content fragments with contextual metadata, such as timestamps and file permissions, enabling multimodal inference from fragmented data while explicitly excluding label-derived features to prevent leakage. PatternMiner is evaluated on established forensic benchmark datasets, including Digital Corpora and AFF4 forensic containers, which simulate realistic fragmentation scenarios. All experiments are conducted under an explicit leakage-controlled evaluation protocol with group-aware data partitioning to ensure that performance reflects generalization to unseen data. Results show that the proposed framework achieves strong performance, with an accuracy of 92.1% and a macro-averaged F1-score of 92.1% under complete input conditions. Furthermore, the model demonstrates resilience to degraded and partially corrupted inputs, including truncation, byte removal, shifting, and fragment reordering. These findings indicate that PatternMiner effectively captures structural and contextual patterns in fragmented data, providing a practical step toward more reliable and data-driven forensic analysis. By combining multimodal learning with rigorous evaluation practices, the proposed framework contributes to developing scalable and generalizable solutions for modern digital forensic environments.

Computers doi: 10.3390/computers15060353

Authors: Ilyas Potamitis

Effective dementia care is often hindered by fragmented communication among patients, informal caregivers, and clinicians. To address this, we introduce an ambient assisted living (AAL) framework designed to establish a continuous, virtual, and unobtrusive connection between an elder’s home and external guardians or medical staff (virtual rounds). The system enables guardians to communicate directly within the home environment, without requiring the older adult to manually accept calls or activate the connection using wearable devices, buttons, or other interfaces. The elders can activate the connection verbally. The structural core of this system relies on three novel hardware configurations designed for zero-interface operation: a remote audio announcement device, a bidirectional intercom, and a “zero-interface mirror” enabling stream-only, real-time video co-presence between patients and guardians. Crucially, the system utilizes a privacy-preserving, staged edge-AI architecture to process data. By default, it operates without long-term persistent storage, selectively transmitting abstracted audio-based behavioral metrics to a secure dashboard. For advanced dementia stages, the system employs ephemeral data retention—specifically a highly restricted, 24 h rolling audio buffer—allowing authorized guardians to verify acute events without permanently exfiltrating raw data. We evaluate this infrastructure through a 10-month longitudinal, single-home feasibility deployment, augmented with historical verified fall data to rigorously test the detection of rare acute events. The study validates the framework’s technical viability, system uptime, and privacy-first architecture in continuously tracking long-term proxy behavioral indicators under real-world conditions. Rather than asserting generalized clinical efficacy, this work demonstrates the operational feasibility of a novel, affordable, technical blueprint for dignified, remote digital care coordination.

Computers doi: 10.3390/computers15060352

Authors: Faisal A. Alshuwaier Fawaz A. Alsulaiman

Detecting fake news is a challenging task and an important area of research for social media researchers. This task also involves clarifying accountability mechanisms that demonstrate the credibility of quotable sources, such as networks that document the spread of misinformation. Deep learning techniques, particularly neural networks that rely on popular graph representation techniques such as graph convolutional networks (GCNs), are increasingly being utilized to detect fake news, fake accounts, and rumors spreading through social media. In this paper, features were extracted using TF-IDF, Bag-of-Words, and bigrams. The evaluation was conducted using the standard Kaggle/ISOT and GossipCop datasets, which include news headlines and published models. Using the extracted features, the proposed GCN-based model/classifier achieved a high detection accuracy of 95% by combining TF-IDF and Bag-of-Words representations. The results demonstrate that the extracted features improve the efficiency of the detection model.

Computers doi: 10.3390/computers15060351

Authors: Zhenqian Wang Yi Dong Meixia Fu Jianquan Wang Jie Sun Qu Wang Yifan Lu Na Chen Ronghui Zhang Wen Zhang

The of personal computer (PC) tasks represents a systems-level challenge that integrates natural language processing, visual perception and mouse–keyboard action control. While existing approaches mainly focus on the application programming interface (API)-based or terminal-based automation, which are incompatible with the majority of applications for the lack of accessible interface. In this article, we propose PCLLM, a novel end-to-end system that automates PC operations by integrating large language models (LLMs) with computer vision techniques to directly control the mouse and keyboard. First, a software knowledge-based prompt engineering method is developed to comprehend software architecture and operational sequences. Second, template matching techniques are integrated for precise element localization, allowing the system to accurately identify and interact. Third, a dual-LLM pipeline is designed to automatically generate the test data, where a questioner LLM generates diverse task commands and the PCLLM executes these tasks, the corresponding process data are recorded automatically for performance evaluation. Finally, PCLLM is further validated through three typically PC applications (Notepad, Wordpad and Calculator), demonstrating its flexible and robust performance towards intelligent PC automation. To evaluate the proposed system, we adopt task completion rate as the primary metric. Experimental results show that PCLLM achieves the highest completion rates of 98.59%, 95.77%, and 52.11% on Notepad for basic, intermediate, and advanced tasks respectively when powered by GPT-4o, outperforming the CogAgent baseline. These results demonstrate the effectiveness of our approach for PC task automation.

Computers doi: 10.3390/computers15060350

Authors: Ireti Hope Ajayi Elena Yuryevna Avksentieva

Behavioural analytics remain a dominant approach for modelling learner engagement and predicting performance in digital learning environments. However, existing approaches are largely retrospective, relying on observable behavioural outcomes rather than modelling the underlying cognitive state dynamics that evolve during sustained learning. This study proposes a nonlinear state-space modelling framework that formalises the interaction between cognitive fatigue, attention, and learning as a continuous-time dynamical system. Fatigue is modelled as a latent state governed by load–recovery dynamics, attention is represented as a fatigue-coupled cognitive resource, and learning accumulation is expressed as an attention-mediated process under saturation constraints. The model is discretised and empirically estimated using time-indexed webcam-derived pilot data (N = 63) and further validated using a large-scale intervention dataset (N = 1245). Parameter estimation is performed using regression-based approximation of the discretised state equations, with cluster-robust inference applied to account for intra-session dependencies. The webcam-derived features were pre-processed using temporal windowing and normalisation to ensure consistency across sessions. The swarm-optimised intervention was implemented through adaptive control of instructional load and recovery scheduling, enabling real-time regulation of fatigue progression. Empirical results demonstrate statistically significant model validity, with fatigue dynamics showing moderate explanatory capability(R2 = 0.543, p<0.001) and attention dynamics also significant (R2 = 0.499, p = 0.004). At the system level, adaptive intervention significantly reduced fatigue and improved learning performance (t(1244) = 14.34, p < 0.001). The findings suggest a transition from retrospective behavioural modelling toward anticipatory cognitivestate regulation, contributing toward a computational foundation for fatigue-aware adaptive learning systems.

Computers doi: 10.3390/computers15060349

Authors: Sibananda Behera Namita Panda Sudhansu Shekhar Patra

A Software-Defined Network (SDN) renders flexible traffic engineering, but consumes a lot of energy. There is an overhead on the control-plane because flow rule updates are always performed and there is energy consumption by the forwarding hardware. Current energy-aware SDN methods mostly focus on Static or Greedy optimizations. This can cause too many Ternary Content-Addressable Memory (TCAM) updates and unstable rule churn when traffic changes over time. This article introduces a Dynamic Flow Rule Placement (DFRP) framework for real-time energy optimization in SDN. It reduces network energy usage, TCAM update costs, and rule churn all at the same time. The suggested framework uses a convex relaxation method to take decisions on binary switches, links, and rule placement. It also uses a minimum-edit round scheme that only allows small rule changes between time slots. To further reduce instability in the control plane, batch scheduling and receding horizon optimization (RHO) techniques are used. The system uses predicted traffic for future time slots to make decisions, but only the actions for the current time slot are executed. The experiments are carried out on two real-world dynamic SNDlib topologies such as Germany50 and Nobel-Germany, using 288 five-minute traffic matrices over a one-day period. Comparative results against Static and Greedy baselines show that DFRP saves approx. 30% energy while cutting down on TCAM update overhead and rule churn by approx. 20%, consistently across both the networks. Hence DFRP can be applied on dynamic traffic large-scale networks for stable and energy-efficient SDN operations.

Computers doi: 10.3390/computers15060348

Authors: Alparslan Sari Ismail Butun

This study investigates product abuse, reconciliation challenges, and cybersecurity risks in warehouse management systems (WMS) within increasingly digitized supply chain environments. As warehouses evolve into data-driven operational hubs, vulnerabilities such as data manipulation, insider threats, and fraudulent activities pose significant risks to financial accountability and system integrity. To address these challenges, this research proposes a security-centric WMS framework that integrates blockchain-based immutable logging, Internet of Things (IoT)-enabled tracking, and artificial intelligence (AI)-driven anomaly detection. The methodology follows a hybrid iterative–incremental development approach, supported by real-world deployment of a prototype WMS implemented using a scalable microservices architecture. Over a five-year operational period, the system processed more than 10 million transactions with no recorded successful cybersecurity incidents leading to data breaches, operational compromise, or unauthorized system access, while achieving improvements in reconciliation accuracy, operational efficiency, and fraud detection capabilities. Results demonstrate reductions in manual reconciliation efforts, mispricing incidents, and operational losses, while maintaining high system availability and low latency. In addition, the reported 18–22% improvement associated with AI-assisted anomaly detection is presented as a simulation-based projection rather than a production-validated measurement. The findings indicate that combining secure software engineering practices with automation, auditability, and advanced analytics can significantly enhance transparency and resilience in warehouse operations. The study concludes that integrating decentralized and intelligent technologies provides a viable pathway toward secure, privacy-preserving, and abuse-resistant warehouse ecosystems.

Computers doi: 10.3390/computers15060347

Authors: Khalid Almohammadi

Group projects are considered a fundamental component of higher education, as they enhance students’ competencies and problem-solving abilities within professional learning environments. Therefore, ensuring student success and providing effective supervision is essential. However, this remains a challenging task due to the reliance on supervisors’ expertise and the diverse characteristics and backgrounds of student groups. In this paper, we introduce a novel theoretical and practical interval type-2 fuzzy logic system (IT2FLS) for early prediction and guidance for novice supervisors by correlating and learning expert supervisors’ assessments according to the required level of support and the risk of failure for student groups needing early intervention. Experimental evaluation was performed based on assessments of 33 graduation projects conducted by expert supervisors, which served as the input–output data for developing interpretable white-box models that allow both novice and expert supervisors to transparently analyse reasoning processes and outcomes. The results demonstrate that the developed IT2FLS predicts the required level of support and the risk of failure for student groups with lower average error and standard deviation, outperforming the encountered Type-1 fuzzy logic systems. This study thus indicates the IT2FLS’s effectiveness in handling linguistic and numerical uncertainties in supervisors’ evaluations of students’ required early interventions.

Computers doi: 10.3390/computers15060346

Authors: Zhanibek Kozhirbayev Assel Omarbekova

The increasing adoption of data-driven educational systems requires reliable methods to predict student readiness for future coursework and support personalised learning pathways. This study proposes a graph-enhanced ensemble framework that integrates curriculum structure and skill-gap awareness to estimate student course readiness. A global prerequisite directed acyclic graph (DAG) of university subjects was constructed to model curriculum dependencies, from which structural features including the PageRank, in-degree, out-degree, and prerequisite chain depth were derived. In parallel, a domain-informed skill cluster mapping grouped subjects into nine interpretable competency domains to enable skill-gap analysis. These curriculum-aware features were combined with academic history, behavioural engagement, and demographic indicators to produce 38 engineered features for each student–subject pair. Six models (CatBoost, XGBoost, LightGBM, FT-Transformer, MLP and TabPFN) were trained and combined using a weighted ensemble. Experiments on a real-world institutional dataset containing 20,581 students and 727,168 records achieved an AUC of 0.8908 for predicting course success. Ablation experiments demonstrate that graph-derived and skill-cluster features provide modest but statistically significant incremental value. The resulting model was integrated into a prototype personalised recommender that prioritizes curriculum-consistent learning pathways. The proposed framework provides an interpretable and curriculum-aware approach for personalised learning. While the model demonstrates strong overall performance, a moderate gender disparity in the false positive rate was observed. Results were obtained on a large longitudinal dataset from a single university, and external validation at other institutions is needed to confirm generalizability.

Computers doi: 10.3390/computers15060345

Authors: Aidana Karibayeva Oleg Myssov Balzhan Abduali Dina Amirova Adina Karybayeva

In this paper, we first present a multimodal architecture called CrossAttn-v1. This model is designed to recover punctuation marks in Kazakh and combines contextual XLM-RoBERTa-large text embeddings with the Whisper large-v3 encoder states via a cross-attention mechanism. In addition, a 4-dimensional prosodic vector and a CRF output layer are used. The model was trained using an adapted Whisper ASR model on 33,332 utterances from the KazakhTTS2 corpus. After adaptation, the word error rate decreased from 45.7% to 4.25%. On the in-domain test set (56,396 tokens), CrossAttn-v1 achieved F1-macro = 0.8485 for recovering five-class punctuation marks. Furthermore, CrossAttn-v1 outperformed the GPT-4o zero-shot model by +0.294 F1 and the M3 Hybrid model based on prosody alone by +0.070 F1. The class analysis showed that the Whisper encoder states were particularly useful for prosody-dependent punctuation. For example, it outperformed M3 Hybrid by +9.5 percentage points on the QUESTION mark and by +20.2 percentage points on the EXCLAIM mark. On 883 out-of-domain natural speech recordings, the model performed similarly to the text-only baseline model (Δ = −0.041, not significant), suggesting that domain mismatch in the Whisper training corpus was a major factor limiting generalization.

Computers doi: 10.3390/computers15060344

Authors: Huaiwen He Miaoling Liu Chenghao Zhou Hong Shen Hui Tian Shuqing Huang

This paper addresses the problem of throughput utility maximization in a non-orthogonal multiple access (NOMA)-enabled wireless power transfer mobile edge computing (WPT-MEC) network with dynamic task arrivals and user cooperation. To promote fairness and effectively handle random task arrivals and time-varying channels, we model the system utility as a nonlinear function of time-averaged throughput. We then formulate a stochastic optimization problem aimed at maximizing utility while strictly maintaining sensor queue stability. By leveraging the Lyapunov optimization framework, the long-term network-wide utility maximization is decomposed into efficient, slot-wise convex subproblems that operate online without requiring prior knowledge of future task arrivals or channel states. We develop a Dynamic User Cooperation Throughput Maximization (DUCTM) algorithm that enables adaptive resource allocation and cooperative computation offloading in an online manner. Theoretical analysis establishes a provable [O(1/V),O(V)] trade-off between utility optimality and queue backlog. Extensive simulations demonstrate that our approach consistently outperforms baseline methods, providing robust and stable performance even under bursty traffic and highly dynamic environmental conditions.

Computers doi: 10.3390/computers15060343

Authors: Guillermo Vera-Amaro José Rafael Rojano-Cáceres

Web accessibility remediation using large language models (LLM) has recently gained attention; however, most approaches remain tool-centric and lack formal architectural grounding. This article introduces a formally structured conceptual model for born-accessible web remediation using LLMs. The model was derived through a systematic literature review and refined under the Design Science Research Methodology. Unlike patch-based repair strategies, it treats remediation as constrained regeneration, producing accessible content from semantically reorganized inputs. The model defines five core components—input acquisition, intermediate transformation, prompt configuration, generative inference, and output evaluation—and formalizes their interactions and decision mechanisms. A controlled demonstration using multiple LLMs (GPT, Gemini) and automated tools (Lighthouse, Axe, WAVE), complemented by checklist-based structural inspection, was conducted. Results indicate that accessibility improvement depends strongly on architectural structuring of transformation and evaluation sequencing. The formalization advances LLM-driven accessibility remediation from empirical experimentation toward a reproducible, decision-governed generative paradigm, providing a structured foundation for the systematic development of accessibility-oriented architectures, frameworks, and software systems.

Computers doi: 10.3390/computers15060342

Authors: Araceli Olmos-Vallejo Lisbeth Rodríguez-Mazahua Isaac Machorro-Cano José Antonio Palet-Guzmán Giner Alor-Hernández Jair Cervantes José Luis Sánchez-Cervantes

The use of open-ended survey questions for data collection has increased significantly across various areas, as has the application of machine learning (ML) and natural language processing (NLP) techniques to analyze respondents’ opinions. In this study, we conducted a scoping review of 79 studies that analyze open-ended answers given in surveys. We structured our review around six main criteria: application of supervised learning, unsupervised learning, Supervised Descriptive Rule Discovery (SDRD), open-ended questions, NLP, and opinion comparison. This approach allowed us to identify the most used tasks, algorithms, and technologies in ML and NLP, revealing areas of opportunity and the main future challenges. We based our review on the methodological framework of Arksey and O’Malley and adapted PRISMA for reporting systematic reviews. Our findings suggest that most studies addressing surveys with open-ended questions were published in 2020 and 2022, predominantly focusing on research and health domains.

Computers doi: 10.3390/computers15060341

Authors: Mohsen Vahabi Muhammad Zohaib Seyed-Sajad Ahmadpour Osman Selvi

Multiplication is a fundamental operation in digital signal processing, embedded computing, and nanoscale arithmetic data paths, where area, delay, and energy efficiency are critical design constraints. However, nanoscale multiplier design is challenged by high interconnect complexity, frequent wire crossings, clock-zone synchronization issues, and the rapid growth of area and latency with operand size. Quantum-dot cellular automata (QCA) technology offers a promising post-CMOS platform for compact arithmetic circuit realization through field-coupled computation and transistor-free switching. This paper presents a single-layer QCA-based Dadda Tree Multiplier (DTM) using layout-aware integration of compact half-adder, full adder, XOR, and carry-skip adder modules. The proposed design emphasizes partial-product compression, routing compactness, clock-aware organization, and area-efficient final accumulation. Functional verification is performed using QCADesigner 2.0.3, while energy-related behavior is evaluated using QCADesigner-E under the conventional QCA simulation framework. The proposed DTM consists of 4282 cells and occupies 6.14 μm2. Compared with a recent compact QCA multiplier baseline, the proposed architecture reduces cell count by 59.12% and occupies area by 39.80%, while maintaining competitive clocking latency. These results indicate that layout-aware integration of arithmetic modules can substantially improve the area efficiency of QCA-based multipliers, making the proposed design a compact arithmetic core for future nanoscale embedded and signal-processing systems.

Computers doi: 10.3390/computers15060340

Authors: Ahlam Alghamdi Samia Dardouri

The rapid expansion of Internet of Things (IoT) infrastructures in smart city environments has increased the demand for reliable intrusion detection systems (IDS). However, many existing studies rely on single-dataset evaluations and inconsistent experimental settings, which can lead to overly optimistic performance estimates. In this study, we propose a standardized benchmarking framework for evaluating artificial intelligence-based IDS across heterogeneous IoT datasets, including CIC-IoT 2023, BoT-IoT, and N-BaIoT. Multiple classical machine learning and deep learning models are evaluated under a unified preprocessing pipeline and a consistent evaluation protocol. A hybrid CNN–BiLSTM–Attention architecture is also implemented as a reference model within this framework. While several models achieve near-perfect performance under intra-dataset evaluation, cross-dataset experiments reveal substantial performance degradation and unstable metric behavior under distribution shifts. These results highlight the limitations of dataset-specific optimization and emphasize the necessity of cross-dataset validation for realistic IoT intrusion detection evaluation. All experiments are conducted under a binary intrusion detection setting (benign vs. attack) to enable consistent comparison across datasets. Consequently, the reported results reflect binary detection performance and do not capture attack-type discrimination.

Computers doi: 10.3390/computers15060339

Authors: Maraga Alex Sunday O. Ojo

The fast growth of AI and large-scale industrial compute infrastructures has led to unsustainable increases in energy consumption and greenhouse gas emissions on a global scale, creating serious sustainability issues in today’s modern cloud computing. The proposed hybrid framework called the Hierarchical Clustering Deep Q-Network Carbon-Aware Placement System (HC-DQNCAPS) was developed as a means to combine energy efficient design with carbon-aware deployment strategies to support intelligent, adaptive and environmentally sustainable workload scheduling and resource allocation for industrial computing systems. This framework uses real time metrics of resource utilization (server and network) and information about carbon intensity to improve the distribution of workloads across geographically distributed cloud and hybrid infrastructures through both Hierarchical Agglomerative Clustering (HAC)- and Deep Q-Network (DQN)-based reinforcement learning models. Multi-objective optimization is leveraged to optimize energy usage, carbon emissions and SLA violations while optimizing resource utilization. The HC-DQNCAPS architecture significantly outperformed such work practices as FCFS, Energy-Aware VM Allocation, Carbon-Unaware RL, PPO, DDQN and MADRL Scheduling, with SLA breaches always less than 5%, and with energy utilization consistently reduced by 30–35%, carbon emissions reduced by 25–30% and resource utilization increased by +20%. The model’s significance and stability were demonstrated using both ANOVA and Wilcoxon signed-rank statistical tests to be significant (p < 0.05) at 95% confidence intervals. Overall, the findings show that there is potential for implementing carbon-aware AI methods in order to maintain economic viability for all computing systems involved in the industrial cloud.

Computers doi: 10.3390/computers15060338

Authors: Stamatios Orfanos Parisis Gallos Christos Panagopoulos Andreas Menychtas Ilias Maglogiannis

Academic stress in higher education is strongly influenced by workload structure and scheduling decisions, yet academic planning sometimes remains static and does not incorporate behavioural or physiological indicators. While existing research focuses on stress measurement and prediction, these approaches are rarely integrated into decision-support mechanisms capable of restructuring academic schedules. This work introduces a Course Digital Twin (CDT) framework that integrates multimodal student data with simulation-based academic planning. The proposed system models course scheduling as a decision-support problem, where alternative configurations are evaluated using a structured stress model combining wearable-derived physiological signals, self-reported stress measures, and contextual academic workload indicators. The framework employs a hybrid approach in which machine learning is used for physiological stress estimation, while schedule adaptation is performed through transparent rule-based mechanisms. The system was implemented as an end-to-end platform including mobile sensing, course configuration interfaces, and instructor analytics dashboards, and was evaluated through a pilot deployment across multiple postgraduate courses. Preliminary results indicate that simulation-based schedule adjustments are associated with reductions in projected peak stress levels and improved workload distribution patterns. The findings demonstrate the feasibility of integrating multimodal stress modelling and Digital Twin simulation into academic planning workflows. The proposed framework provides a foundation for future stress-aware scheduling systems, although further large-scale validation is required to establish its effectiveness and generalizability.

Computers doi: 10.3390/computers15060337

Authors: Shujing Tong Yongfei Wu

Automatic video anomaly detection remains challenging because abnormal events are infrequent, visually heterogeneous, and weakly bounded in time. This study proposes an unsupervised framework trained only with normal video segments. The framework integrates sliding-window segment construction, dual-view perturbation, a two-branch spatio-temporal encoder, exponential moving-average prototype updating, prototype-guided contrastive optimization, and a robust anomaly score composed of prototype deviation, second-order temporal residual, and local-neighborhood sparsity. Experiments were conducted on UCSD Ped2, CUHK Avenue, and ShanghaiTech under the same input size, segment length, optimizer, and threshold protocol. The proposed model achieved AUC values of 97.4%, 91.8%, and 83.7% on the three datasets, respectively, with an average AUC of 91.0% and an average F1 score of 88.1%. Relative to the baseline contrastive model, the average AUC increased by 2.4 percentage points, and the average F1 score increased by 2.8 percentage points. Across three independent runs, the improvement over the contrastive baseline was statistically significant (paired two-sided t-test, p = 0.018). Ablation and sensitivity analyses indicate that the performance gain is mainly attributable to spatio-temporal joint encoding, prototype traction, temporal residual scoring, and local-neighborhood support. These results show that contrastive representation learning, explicit prototype updating, and temporal-aware scoring can jointly produce a stable representation of normal behavior without using abnormal samples during training.

Computers doi: 10.3390/computers15060336

Authors: Weigang Wang Lintao Xian Ziyuan Cui

Due to low storage cost and fast query efficiency, cross-modal hashing has attracted considerable interest in multi-modal data retrieval. However, existing hashing methods face several challenges: one major challenge arises from the neglect of both local and non-local neighborhood structural relationships within multi-modal information, which makes it difficult to establish fine-grained semantic consistency associations between heterogeneous modalities. Additionally, the imbalance in the number of training samples limits the improvement of retrieval performance. To address these challenges, a Deep Neighborhood-similarity Preservation Hashing (DNsPH) method is proposed for cross-modal retrieval. To obtain the high-order statistical features of images, we first design a Context-aware Cross-layer Bilinear Fusion Network (C2BF-Net), which uses Long Short-Term Memory (LSTM) to model the context-dependent features of different convolutional layers. Furthermore, the image, text, and semantic labels information are fused through an adaptive weighting strategy to reconstruct the joint semantic similarity matrix to explore the fine-grained neighborhood structure between different modalities. Finally, we introduce a multi-similarity loss based on an adaptive margin to mining and weighting informative sample pairs, to alleviate the impact of sample imbalance on model training, and thereby generate more discriminative hash codes. Extensive experiments performed on the MIRFLICKR-25K and NUS-WIDE datasets demonstrate that DNsPH outperforms state-of-the-art cross-modal retrieval applications.

Computers doi: 10.3390/computers15060335

Authors: Emad-ul-Haq Qazi Muhammad Hamza Faheem Abdulrazaq Almorjan

Phishing sites are increasingly causing harm to consumers, commercial enterprises, and the online infrastructure. Online safety is dependent on how well these evil sites can be detected in time and correctly. A number of solutions that exist are based on lexical, token features, or structural hints. Although useful to a certain degree, these methods tend to lose more contextual meaning in URLs. This paper presents SemanticPhishNet, a hybrid detection system that uses semantic knowledge to detect phishing attacks by utilizing semantic knowledge via a transformer-based system to process HTML documents and URL information to produce accurate and efficient phishing detections. The architecture uses MiniLM (identical type as distillbert) to obtain contextual embeddings of cleaned HTML and augmented text of URLs and a simple dense classifier to perform effective binary classification. A stratified three-way split of data was used to evaluate the model with real-world obfuscation patterns like replacement of “http” by “hxxp”. The experimental findings show that SemanticPhishNet has high performance in various measures, outperforming other state-of-the-art models in accuracy, recall and generalization ability. We conduct experiments on cross-validation and external validation with independent data. The framework exhibits good performance (96–97% cross-validation accuracy) and external evaluation demonstrates realistic generalization (67% accuracy), albeit revealing the difficulties of domain shift in phishing. The proposed model performs better than many of the existing models in the real world. The confusion matrices and ROC analysis indicate that the phishing and benign classes are consistently separated in both the validation and test sets. The findings show that the suggested model is efficient, stable, and scalable to the present-day phishing detection. The paper stresses the importance of appropriate evaluation techniques, such as leakage-aware splits and cross-dataset evaluation.

Computers doi: 10.3390/computers15060334

Authors: Muharman Lubis Paxilla Chairany Alif Noorachmad Muttaqin Arif Ridho Lubis

Enterprise systems are widely adopted in organizations, yet user acceptance remains a major challenge due to the complex interplay of cognitive, social, motivational, and innovation-related factors. Existing technology acceptance models often provide fragmented explanations by focusing on limited determinants. This study proposes TAM 4, an exploratory framework integrating constructs from the Technology Acceptance Model (TAM), Unified Theory of Acceptance and Use of Technology (UTAUT), Hedonic-Motivation System Adoption Model (HMSAM), and Diffusion of Innovation (DOI). The study was conducted in the context of enterprise application usage and professional enterprise system training environments involving organizational users, trainees, and practitioners. Data were collected from 115 enterprise system users (trainees and practitioners). To consolidate overlapping indicators and strengthen construct definition, principal component analysis (PCA) was applied, yielding seven higher-order constructs that explain 81.642% of cumulative variance. The framework was validated using PLS-SEM with three scenario-based structural models (full mediation, partial mediation, and direct effects). The results show that Model 3 provides the best fit and predictive performance (SRMR = 0.048; NFI = 0.786), indicating that enterprise system adoption is better explained through a direct effect structure rather than a purely mediated TAM pathway. The novelty of this study lies in introducing TAM 4 as a PCA-driven multi-theory acceptance model and evaluating its explanatory robustness through multi-scenario model comparison, offering practical insights for improving enterprise system implementation strategies.

Computers doi: 10.3390/computers15060333

Authors: Makbal Baibulova Ayagoz Mukhanova Aliya Abdukarimova Lazzat Abdykerimova Bulat Serimbetov Madi Akhmetzhanov Zhanat Seitakhmetova Elmira Yeshtayeva Murizah Kassim Aizat Amirbay

Objective and reproducible diagnosis of attention-deficit/hyperactivity disorder (ADHD) remains challenging because of the limited availability of reliable electroencephalography (EEG) biomarkers and the high variability of neural signals. This study proposes a computational framework for ADHD detection based on dynamic functional connectivity and network-index modeling. Multichannel EEG recordings were transformed into temporal connectivity graphs using sliding-window correlations of band-limited amplitude envelopes. Several network-index models were evaluated, including linear, graph-based, recurrent, and hybrid spatio-temporal approaches. The proposed Hybrid Spatio-Temporal Graph Transformer demonstrated moderate, yet reproducible, subject-level classification performance. On the independent test set, the model achieved an accuracy of 63.16%, a balanced accuracy of 62.22%, a sensitivity of 80.00%, a specificity of 44.44%, an F1-score of 69.57%, and an AUC-ROC of 0.7444. Additional analysis of the derived network index demonstrated moderate intergroup separability, with a mean index shift of 1.16, Cohen’s d = 0.73, Pearson’s r = 0.36, and distribution overlap = 0.72. These findings suggest that the proposed framework captures informative spatio-temporal EEG connectivity patterns associated with ADHD; however, the model’s diagnostic applicability should be considered preliminary and requires validation in larger independent cohorts.

Computers doi: 10.3390/computers15060332

Authors: Ganga Sagar Soni Abhinav Shukla R Kanesaraj Ramasamy Pritendra Kumar Malakar Parul Dubey

Thyroid diseases are very common endocrine diseases that afflict millions of people around the world and need proper and timely diagnosis to ensure proper treatment. Although machine learning and hybrid metaheuristic methods have advanced, current models have high computation costs, low interpretability, and low probability calibration, which limit their use in clinical settings. In this research, a new DFSel-FT (Differentiable Feature Selection and an FT-Transformer) system is suggested, which combines DFSel-FT to allow one to diagnose thyroid disease effectively and interpretably. It employs Concrete (Gumbel-Softmax) gates to select the features end-to-end to make sure that only the most relevant clinical attributes are carried through the training. A Transformer-based architecture is then used to process the chosen features to learn intricate interdependencies. The model is trained with class-balanced focal loss and temperature scaling to better enhance calibration. Experimental evaluation on the UCI Thyroid Disease Dataset (22,632 samples) showed that the proposed model achieved 97.85% accuracy, 97.65% Macro-F1, and 98.10% AUC-OVR, showing competitive performance compared with traditional machine learning models, modern tabular deep learning baselines, and hybrid metaheuristic methods. Other indicators of robustness and reliability include MCC (0.955), Cohen Kappa (0.951), and small calibration error (ECE = 0.021). SHAP and LIME explainability analysis reveals clinically relevant features that include TSH, TT4, and T3. The proposed framework provides a balanced integration of predictive performance, interpretability, and probability calibration, making it a promising benchmark-level framework for interpretable and calibrated thyroid disease classification, requiring external clinical validation before real-world deployment.

Computers doi: 10.3390/computers15060331

Authors: Robert Campbell

Anthropic’s April 2026 Claude Mythos Preview release established a new operational threat category: frontier AI systems whose extended-context reasoning, recursive self-correction, native system-tool integration, and agentic scaffolding render dominant AI safety paradigms—RLHF, output filtering, contractual access vetting, human-in-the-loop supervision—insufficient as sole controls. This paper develops a defense-in-depth reference architecture against that category, structured around four named contributions: a five-indicator operational definition of the Mythos-class (capability conjoined with scaffold, access pattern, autonomy depth, and persistence); the Mythos-Class Posture Rubric (MCPR), a three-tier detection framework spanning evaluation, deployment, and runtime with explicit routing to mitigation layers; a four-layer mitigation stack comprising the Vetted-Access Operational Pattern (VAOP), Authority-Bound Output Release (ABOR) cryptographically grounded in FIPS 203/204/205 post-quantum primitives, and the Compute-Plane Isolation Profile (CPIP); and an integrated architecture that crosswalks to the NIST AI Risk Management Framework, NIST Cybersecurity Framework 2.0, and CISA Zero Trust Maturity Model 2.0. The architecture is applied to three deployment surfaces—post-quantum cryptography migration, federal AI supply-chain assurance, and critical-infrastructure operational technology defense—demonstrating that the four contributions generalize across heterogeneous operational contexts. The contribution is a reference design rather than a deployed system; limitations, falsifiability criteria, and a research agenda for empirical refinement are developed.

Computers doi: 10.3390/computers15060330

Authors: Deniz Poyraz Gök Nuri Kara

This study examines recent research trends in the use of serious games for science education through a bibliometric analysis of 340 articles and a qualitative content analysis of 56 studies published between 2020 and 2025 in the Web of Science Core Collection. By combining these approaches, the study provides a comprehensive view of both research patterns and how serious games are designed and used in science education. The findings indicate that the field is maturing, with research moving beyond general effectiveness toward understanding how serious games support learning in different contexts. Most studies report positive effects compared to traditional instructional methods. However, results vary across contexts and depend on factors such as design, implementation, and learner characteristics. Research is mainly focused on higher education and is largely driven by leading countries such as the USA and China, although participation from developing countries is increasing. The growing use of immersive technologies, such as augmented and virtual reality, offers new opportunities for interactive and multimodal learning but may also increase cognitive load in certain contexts. There is also growing interest in non-digital games, which have received limited attention despite their effectiveness. Overall, the findings show that more systematic research and clearer design frameworks are needed to better understand how serious games can be used in science education.

Computers doi: 10.3390/computers15060329

Authors: Lina Zhao Hua Huang Ning Li Yunxiao Wang Ming Li

With the rapid evolution of cyberattack techniques, identifying dynamic behavioral intents from Application Programming Interface call sequences has become a fundamental modality for ensuring reliable malware detection and information security. However, existing detection methods face the dual challenges of semantic sparsity and inadequate spatial dependency modeling when processing these sequences, which fundamentally undermines their stability against complex structural variations and in-the-wild evasive patterns. To address these critical vulnerabilities, we propose LLM-SGCF, a highly effective malware detection framework that jointly models deep behavioral semantics and spatial structures. Specifically, our framework leverages generative Large Language Models, which are subsequently encoded by BERT, to transform sparse API calls into rich and contextualized descriptions. Concurrently, it employs a novel Spatially Guided Convolution (SGC) module to localize critical malicious segments and extract cross-position dependencies in a two-dimensional semantic space. Extensive experiments on the public Aliyun and Catak datasets demonstrate that LLM-SGCF exhibits exceptional resilience to real-world structural complexity and significantly outperforms state-of-the-art baselines, achieving a peak binary-classification accuracy of 95.82%. Further ablation analyses confirm that the synergistic fusion of semantic enhancement driven by Large Language Models and spatial structural modeling dramatically improves the resilience of the framework against complex attack chains, providing a highly reliable paradigm for next-generation malware recognition systems.

Computers doi: 10.3390/computers15050328

Authors: Weiqi Wang Gwo-Chin Ching Soo Fun Tan

The Internet of Vehicles (IoV) is rapidly emerging as a core component of intelligent transportation systems, enabling real-time communication among vehicles, infrastructure, and cloud platforms. However, the increasing interconnectivity of vehicular systems and the advancement of quantum computing introduce significant security challenges to existing cryptographic mechanisms. Conventional schemes such as RSA and Elliptic Curve Cryptography (ECC) are vulnerable to quantum attacks and are computationally inefficient for resource-constrained vehicular environments. To address these limitations, this paper proposes a Double-Ring Hybrid Sparse NTRU (DRH-SNTRU) framework, a lightweight and quantum-resistant cryptographic scheme for secure IoV communication. The proposed framework introduces three key enhancements: (i) controlled-support sparse polynomial structures to reduce polynomial multiplication complexity while improving entropy distribution; (ii) a double-ring algebraic architecture that separates key operations from message processing to enhance structural security and minimize coefficient leakage; and (iii) hybrid ephemeral keys derived from contextual entropy to strengthen forward secrecy and adaptive security. An optional ciphertext evaluation mechanism is further incorporated to detect malformed and replayed ciphertexts prior to decryption. Security analysis demonstrates that the proposed framework achieves IND-CPA security under the hardness assumption of the NTRU lattice problem and can be extended to resist chosen-ciphertext attacks through the integrated validation mechanism. Experimental benchmarking across polynomial dimensions N = 512 to 8192 demonstrates that DRH-SNTRU achieves low setup overhead below 3 μs, efficient decryption latency of approximately 305.64 μs at N = 8192, and compact sparse private key representation of only 117 bytes at higher dimensions. Compared with Standard NTRUEncrypt, NTRU-HRSS, and Ring-LWE Encryption, the proposed framework demonstrates improved decryption efficiency, lightweight storage overhead, and enhanced ciphertext integrity protection while maintaining practical scalability for resource-constrained post-quantum IoV environments.

Computers doi: 10.3390/computers15050327

Authors: Orazmukhamed Bekmurat Darkhan Akpanbetov Ainur Tursynkhan Laura Demeubayeva Zhansaya Duisenbekkyzy Kanibek Sansyzbay Shingis Kadirkulov Yelena Bakhtiyarova

The timely detection of psycho-emotional risks has become increasingly important due to the rapid growth of social media platforms. This study examines user-generated text as a potential source of early indicators of psychological vulnerability. The proposed NLP-based framework incorporates behavioral features to improve the interpretation of users’ psycho-emotional states. In addition to text classification, the study considers structured behavioral indicators to support psycho-emotional risk analysis. Particular attention is given to interpretability. SHAP-based techniques are applied to reveal the contribution of individual features and to provide a clearer explanation of model predictions. The evaluation was conducted on publicly available datasets containing textual data and aggregated behavioral/physiological indicators. No raw physiological streams, wearable sensor data, or biometric recordings were used. The two datasets were employed in complementary experimental settings and were not aligned at the individual-sample level; accordingly, the broader analytical perspective explored in this study should not be interpreted as a single end-to-end or fully aligned multimodal learning framework. The proposed BERT-based model with SHAP interpretability achieved an accuracy of 96.3%, an F1-score of 0.96, and a ROC–AUC score of 0.98, showing consistent improvement over baseline models, including Random Forests and Support Vector Machines.

Computers doi: 10.3390/computers15050326

Authors: Waleed K. Abdulraheem Emad Mohammed Ibbini Hasan Kanaker Sami Smadi Nader Abdel Karim Hussam N. Fakhouri Layla Albdour Sandi Fakhouri

Role-Based Access Control (RBAC) is the de-facto mechanism for preserving Kubernetes and other cloud-native container platforms, however real deployments occasionally drift away from the principle of least privilege as clusters, teams, and services improve. This paper introduces an automated RBAC hardening framework that formulates least-privilege policy design as a limited optimization problem over RoleBindings and ClusterRoleBindings. The objective combines (i) a permission-risk score for namespaced and cluster-scoped actions with (ii) an operational complexity term that discourages overly large binding sets. Solid limitations encode functional requirements as well as practical security policies, which includes namespace allowlists, role scoping rules, administrative restrictions on cluster-wide bindings, binding budgets, and separation-of-duty requirements expressed by utilizing capability classes. To allow optimizer-agnostic search while protecting Kubernetes RBAC semantics, we analyze candidate policies by utilizing a unified penalty-based fitness function that compines risk, complexity, and constraint violations into a single scalar value. We utilized ten metaheuristic as a benchmark including baseline search paths on a Kubernetes-inspired instance and report feasibility and least-privilege quality metrics (precision, recall, F1, and over-privilege ratio) parallel to RB/CRB counts and excess risk as a structural indicators. Outcomes present that feasibility is the prime challenge, and is restricted to a subset of optimizers reliably arrives to entirely feasible and compact arrangements within the exact budget, indicating the practicality of metaheuristic enhancement for systematic RBAC reduction in containerized cloud computing environments.

Computers doi: 10.3390/computers15050325

Authors: Bayan Baz Afraa Attiah Abeer Hakeem Nada M. Almani

Crime prediction can support proactive public-safety planning, but practical deployment also requires outputs that are reliable and explainable. This study proposes XCrime-LLM for next-week crime occurrence prediction, in which engineered spatio-temporal features are serialized into a fixed prompt format and used to fine-tune GPT-4.1-mini to produce schema-guided JSON outputs from New York City Police Department (NYPD) incident records. The proposed XCrime-LLM framework is evaluated against prompting and trained baselines in New York City and further examined for cross-city transfer in Chicago. Supervised fine-tuning improved GPT-4.1-mini compared with the prompting baselines, increasing Micro-F1 from 0.7478 to 0.8095 and Macro-F1 from 0.7485 to 0.8075, while remaining competitive with the trained baselines. In the cross-city evaluation on Chicago, the fine-tuned GPT-4.1-mini outperformed the base GPT-4.1-mini without further fine-tuning or city-specific adaptation, raising Micro-F1 from 0.8277 to 0.8650 and Macro-F1 from 0.8693 to 0.9020. For explainability under black-box access, KernelSHAP identified last28_mean as the most influential feature across all crime types, while targeted ablation provided additional evidence of the model’s reliance on this feature. These findings suggest that the framework supports competitive next-week crime occurrence prediction while remaining explainable under black-box deployment constraints.

Computers doi: 10.3390/computers15050324

Authors: Xiaodong Xing Zhifeng Zhang Bo Wang

Edge server placement (ESP) is a critical determinant of service quality in edge–cloud computing systems, yet existing solutions often neglect the inherent collaboration between edge and cloud, leading to suboptimal performance under dynamic workloads. To address this gap, this paper proposes a novel meta-heuristic edge server placement algorithm based on the Coati Optimization Algorithm (COA). We first formulate the ESP problem as a constrained binary nonlinear programming model that explicitly incorporates edge–cloud collaboration, aiming to minimize the average request processing delay. The proposed COA-based solver features a compact one-dimensional encoding scheme that simultaneously represents server placement and request offloading decisions, a tailored boundary correction mechanism to enforce coverage and atomicity constraints, and a balanced exploration–exploitation strategy inspired by coatis’ natural hunting and escape behaviors. Extensive simulations are conducted, comparing the proposed algorithm against ten representative heuristic and meta-heuristic algorithms, including GA, PSO, DE, GWO, and their variants. The experimental results demonstrate that our algorithm significantly outperforms all compared methods in terms of the mean, minimum, and standard deviation of the overall average processing delay. Specifically, it achieves a 98.2% reduction in the mean delay relative to suboptimal algorithms while maintaining near-zero variance, confirming its effectiveness, efficiency, and robustness. The proposed algorithm provides a promising solution for service providers to enhance quality of service through optimal edge server deployment and request offloading under edge–cloud collaboration.

Computers doi: 10.3390/computers15050323

Authors: Shen Shi Weiji Yu Ruifeng Yu Yizhuo Zhang

Single image super-resolution (SR) is an important part of image processing, which aims to improve the spatial resolution of images. This is a typical ill-posed inverse problem. The main difficulty is that a low-resolution image block usually corresponds to multiple high-resolution image blocks. The existing methods cannot provide enough correlation to determine the unique high-resolution image block, which leads to artifacts and image distortion in the reconstructed image. To address this problem, a method (EHNet) is proposed to achieve super-resolution by using a Hybrid-Channel Fusion Block (HCFB) and an Enhanced Dual-Convolution Block (EDCB). The EDCB effectively enhances the network’s ability to capture image details and textures by combining local and global feature processing. The HCFB strengthens the information interaction between channels by combining channel segmentation with large-kernel convolution, fully explores feature dependencies, and thus optimizes the feature extraction effect. Experimental results show that the super-resolution reconstructed image of EHNet achieves 32.59 dB PSNR and 0.9006 SSIM on the Set5 ×4 SR benchmark, outperforming several state-of-the-art SR methods. In addition, the model exhibits notable improvements in artifact suppression, and the reconstructed image’s subjective visual impact surpasses that of other current techniques.

Computers doi: 10.3390/computers15050322

Authors: Murray Eugene Jennex Abraham Abby Sen Jeen Mariam Joy

This paper presents a framework for integrating Artificial Intelligence (AI) into Knowledge Management (KM), using the Jennex–Olfman KM Success Model as a foundation. Through a literature review and a thematic analysis of 400 practitioner comments from the global SIKM Leaders community, the study examines how AI is being applied in KM and the implications for practice. Findings highlight that AI expands KM across diverse sectors, enhances efficiency through automation and workflow integration, and supports human judgment in knowledge tasks. At the same time, risks concerning bias, accuracy, transparency, governance, and infrastructure remain central challenges. Mapping these insights to the KM Success Model shows that AI strengthens system and knowledge quality while requiring leadership and governance to safeguard service quality. The analysis extends the model by extending construct definitions with AI and moderating all constructs with AI. Overall, the study concludes that AI can and should be integrated into KM. Successful AI integration is best understood not as isolated technical interventions, but as extensions of KM success theory.

Computers doi: 10.3390/computers15050321

Authors: Theocharis Tsenis Vasileios Barmpagiannos Evangelos D. Spyrou Vassilios Kappatos

The integration of quadruped robotics with advanced sensing technologies offers a transformative approach to forest management, particularly for real-time measurement of tree Diameter at Breast Height (DBH). This paper introduces a novel methodology by deploying a quadruped robot equipped with GPS, LiDAR, and an aligned high-definition camera to patrol forest paths via a developed dynamic autonomous mission. Utilizing a YOLO-based model for trunk detection, the methodology retrieves precise DBH measurements and corresponding geotags, constructing a spatial database of DBH ground-truth data. This database serves as a real-time ground-truth lookup table to calibrate allometric equations used in drone-based crown detection missions, enhancing the accuracy of forest biophysical attribute estimations such as tree height, volume, and biomass. Experimental validation demonstrates high precision in DBH estimation (error < 5% in controlled tests), supporting automated, around-the-clock data collection for sustainable forest management in Mediterranean ecosystems.

Computers doi: 10.3390/computers15050320

Authors: Miguel Ángel González Maestre Javier Cubero Juánez Alejandro de la Hoz Serrano Lina Melo

Automated short-answer grading (ASAG) in educational contexts faces a fundamental trade-off between predictive performance, interpretability, and methodological transparency, particularly under data-constrained educational settings. While recent approaches rely on deep learning architectures, these models require large annotated datasets and offer limited transparency, restricting their applicability in authentic classroom environments. This study proposes a fully specified and interpretable machine learning pipeline for ASAG across multiple educational concepts. The approach is based on a shared TF–IDF representation and evaluates three linear classifiers—Logistic Regression, Multinomial Naïve Bayes, and Linear Support Vector Machines—under a stratified cross-validation framework adapted to small datasets. Model performance is assessed using accuracy, precision, recall, and F1-score. Statistical comparisons using the Wilcoxon signed-rank test indicate exploratory evidence of statistically significant differences between classifiers, although the observed differences remain small in practical magnitude. Additionally, the methodology incorporates token-level analysis to identify discriminative lexical patterns and examine consensus across classifiers. To enhance interpretability, tokens are presented using a bilingual Spanish/English representation while preserving the original feature space. The results across ten concept-specific datasets show consistent performance across models (accuracy ≈ 0.82–0.88) and reveal stable lexical patterns consistently associated with model predictions of correctness. The findings highlight that lightweight, interpretable models can provide consistent and reliable performance under resource-constrained educational conditions. The proposed framework contributes a stability-oriented and interpretable evaluation paradigm for ASAG, offering a practical alternative to data-intensive approaches in educational assessment. It is intended as a methodological reference protocol rather than a performance benchmark. The findings should be interpreted as evidence of within-context consistency instead of broad external generalization.

Computers doi: 10.3390/computers15050319

Authors: Muntakimur Rahaman Azwan Mahmud Azlan Abd Aziz Osama M. S. Abujawa Ji-Jian Chin

Recent studies suggest that few-shot and zero-shot learning methods, drawing on meta-learning, self-supervised approaches and metric-learning ideas, can classify encrypted traffic (TLS 1.3 and QUIC) with competitive accuracy across different protocol conditions. This systematic literature review (SLR) investigates 22 studies selected from an initial pool of 500 papers using PRISMA 2020, focusing on current methodologies for non-stationary network traffic classification, with particular attention to few-shot, zero-shot, and meta-learning approaches. The research addresses four questions: (1) Which approaches have been employed for non-stationary network traffic classification and threat detection? (2) How do hybrid or cross-domain models improve adaptation, detection and overall efficiency? (3) What benchmarking standards exist for the datasets and evaluation metrics in use? (4) How do these methods address concept drift? This review identifies a range of approaches for capturing and analysing non-stationary network traffic but also reveals a significant gap in the empirical evidence addressing the last two questions. This points to a need for targeted experiments on continuously evolving network traffic and zero-day polymorphic attacks, both of which are central to the development of the next-generation adaptive intrusion-detection framework.

Computers doi: 10.3390/computers15050318

Authors: Jing Wang Quan Zhou Anyi Yang Junyu Lin

Open-vocabulary semantic segmentation (OVSS) aims to achieve pixel-level object segmentation guided by arbitrary natural language descriptions. Although pre-trained vision–language models (VLMs) have significantly advanced the development of OVSS, their reliance on the Vision Transformer (ViT) architecture imposes a fundamental constraint on dense prediction. Specifically, the absence of hierarchical downsampling in ViT-based VLM results in single-scale representations that trade spatial localization for global semantics. To address these issues, this paper proposes a hierarchical boundary-constrained decoding network for OVSS, called CLIP-HBD. Our approach leverages VLM semantic priors to reconstruct multi-scale features and introduces a boundary-constrained decoding strategy to refine edge details. Specifically, CLIP-HBD leverages a ConvNeXt-based backbone alongside a hierarchical adaptation mechanism to fuse multi-layer VLM features, generating a comprehensive multi-scale representation. To address the issue of boundary inaccuracy, we perform explicit boundary prediction based on multi-scale representations, where the resulting boundary maps are subsequently transformed into structural constraints to steer the decoder’s focus toward boundary regions. By integrating structural constraints with hierarchical features, the decoding process effectively maintains semantic consistency and restores precise object boundaries. Extensive experiments demonstrate that CLIP-HBD achieves superior performance in both segmentation precision and boundary quality across multiple benchmarks.

Computers doi: 10.3390/computers15050317

Authors: Hikaru Fukuda Masaki Morishita Kosuke Muraoka Shino Maeda Taiji Nakamura Manabu Habu Shuji Awano Kentaro Ono

Large language models (LLMs) are increasingly used in professional examinations, but their relative performance on dental board-style questions remains unclear. This study compared two reasoning-optimized models, GPT-o3 and GPT-5T, with a general-purpose multimodal model, GPT-4o, using 399 Japanese dental board-style multiple-choice questions from 2018 to 2022. All questions were presented in Japanese, and items originally accompanied by charts, photographs, or other figures were analyzed separately from items without visual materials. Accuracy and item-level agreement were assessed using pairwise McNemar tests, stratified analyses according to the original presence of visual materials, the Breslow–Day test for homogeneity of odds ratios, and two-proportion z-tests. GPT-5T achieved the highest overall accuracy (294/399, 73.7%), followed by GPT-o3 (257/399, 64.4%) and GPT-4o (255/399, 63.9%). Pairwise McNemar tests showed that GPT-5T outperformed both GPT-4o (Holm-adjusted p = 0.00098) and GPT-o3 (Holm-adjusted p = 0.00072), whereas GPT-o3 and GPT-4o did not differ significantly (Holm-adjusted p = 0.920). Accuracy was lower for questions originally containing visual materials than for questions without such materials across all three models (GPT-4o: 49.7% vs. 72.2%; GPT-o3: 55.1% vs. 69.8%; GPT-5T: 59.9% vs. 81.8%). The advantage of GPT-5T was more evident in questions without visual materials, and heterogeneity across question formats was observed for GPT-5T versus GPT-o3. GPT-5T showed the strongest performance in this dataset. Questions originally containing visual materials were associated with lower accuracy across all models. Because the comparison was based on distinct item groups rather than experimentally manipulated visual conditions, this result should be interpreted as a difference across question formats and may also reflect differences in item composition and difficulty between the two groups.

Computers doi: 10.3390/computers15050316

Authors: Osman Selvi Seyed-Sajad Ahmadpour Muhammad Zohaib Naim Ajlouni

Edge Internet of Things (IoT) blockchain deployments increasingly rely on continuous transaction ingestion from resource-constrained IoT devices to nearby edge gateways over heterogeneous wireless links. In this setting, transient channel noise and packet corruption can inject invalid payloads into the edge processing pipeline and trigger unnecessary buffering, parsing, and, most critically, computationally expensive cryptographic operations such as digital signature verification. This leads to wasted computation, increased latency, and reduced energy efficiency at the edge, particularly under dense IoT traffic. This paper presents an energy-aware and fault-tolerant Quantum-Dot Cellular Automata (QCA)-based integrity pre-filter for IoT-to-edge blockchain transaction ingestion. At the circuit level, we adapt and modify a previously reported fault-tolerant five-input majority gate (MV5) structure and use it as a robust primitive for nanoscale integrity-screening circuits. Building on this modified MV5, we design a set of QCA integrity blocks, including a parity checker, a compact XNOR gate circuit, a parity-bit generation circuit, and a sender-to-channel/receiver nano-communication integrity workflow suitable for early screening of corrupted payloads. Compared with the best previously reported baseline considered in this study, the modified MV5 achieves 76.47% tolerance to single-cell omission defects, corresponding to a 17.47 percentage-point increase and an approximately 29.61% relative improvement over the prior 59% omission-tolerance result, while preserving 100% tolerance against extra-cell deposition defects. At the system level, the proposed circuit is discussed as a potential early screening stage for edge-IoT blockchain transaction ingestion. A bounded analytical model is used to estimate the possible reduction in unnecessary signature-verification workload under assumed corruption and detection conditions. This analysis is not intended as a deployment-level validation; full edge-node implementation, throughput measurement, queueing-delay evaluation, real traffic traces, retransmission behavior, and empirical signature-verification profiling remain future work. The proposed parity/chunk-parity pre-filter is designed for low-cost detection of random transmission-induced corruption and does not replace cryptographic authentication, hashing, digital signatures, CRC-based detection, or blockchain validation. All proposed designs are validated using QCADesigner tools.

Computers doi: 10.3390/computers15050314

Authors: Saurabh Sanjay Singh Deepak Gupta

Sustainable manufacturing requires schedules that balance environmental responsibility with delivery reliability. This paper studies the Carbon-Aware Flexible Job Shop Scheduling Problem with Tardiness Penalty (CAFJSP-T), where total carbon emissions and total tardiness penalty are the primary objectives. We propose a Policy-based Rough Optimization with a Large Neighborhood Search (Pro-LNS) framework integrating Proximal Policy Optimization (PPO) and adaptive Large Neighborhood Search (LNS). PPO constructs a feasible schedule by selecting operation-machine assignments from job-readiness, machine-availability, earliest-completion, and critical-path features. This policy-generated schedule provides a structurally informed incumbent, enabling LNS to avoid unguided search and focus destroy-and-repair refinement on high-impact operations. Both phases use the same normalized scalarized carbon-tardiness objective, which guides PPO rewards and LNS removal, reinsertion, and acceptance while preserving precedence, eligibility, and capacity constraints. Experiments on small, medium, and large workcenter benchmarks show strong due-date performance and controlled carbon emissions. Under equal objective weighting, Pro-LNS achieves a median optimality gap of 6.12% relative to the exact formulation, with all instances within 14%, while requiring 4.08 s on average and at most 10.51 s. Comparisons with PPO-only, Advantage Actor-Critic (A2C), Soft Actor-Critic (SAC), and Genetic Algorithm (GA) schedulers show that Pro-LNS attains the best weighted scalarized objective across representative instance-weight settings. Friedman and Holm-corrected Wilcoxon tests confirm significant improvements over all competitors, with average weighted-objective gains of 4.90%, 7.25%, 8.81%, and 9.51% over PPO-only, A2C, SAC, and GA, respectively. These results demonstrate that Pro-LNS is an effective and computationally practical hybrid approach for carbon-aware, tardiness-sensitive flexible job shop scheduling.

Computers doi: 10.3390/computers15050315

Authors: Fateme Khanipour Zeinab Shahbazi Sara Behnamian Fatemeh Fogh Nathan Blood

We investigate the linguistic complexity and emotional valence of popular song lyrics across English (n=1491), Spanish (n=307), and German (n=225), using an analytical corpus of 2023 tracks drawn from 2113 deduplicated tracks on Spotify’s weekly Top 200 charts (2019–2021). Transformer-based sentiment analysis is combined with complexity-science tools to characterize both the affective content and the structural organization of commercially successful lyrics. A multilingual BERT model reveals a mild negative skew across all three languages (63.7% negative overall); the 1.003-point English–German gap observed under the English-centric VADER lexicon collapses to 0.127 points under BERT, indicating that earlier cross-linguistic sentiment differences are largely measurement artifacts. Word frequency distributions follow Zipf’s law in all three languages (R2>0.96), with English steepest (α=1.409) and German shallowest (α=1.181). Detrended fluctuation analysis indicates persistent long-range correlations (H≈0.66–0.76; none of the 50 shuffled surrogates exceeded the observed values), and multifractal singularity spectra are statistically indistinguishable across languages once corpus size is controlled (all pairwise Mann–Whitney p>0.13). Streaming counts within the Top 200 are concentrated (German Gini =0.556) but, given the truncated single-snapshot sample, are reported as within-chart descriptors rather than population-level scaling.

Computers doi: 10.3390/computers15050313

Authors: Noha Rashad Mahmoud Hussam Elbehiery Basheer Abdel Fattah Youssef Hanaa Bayomi Ali Mobarz

High-resolution satellite imagery is pivotal for accurate analysis in remote sensing applications, including land-use monitoring, urban planning, and environmental assessment. However, obtaining such data is often costly and limited. Consequently, super-resolution techniques, such as deep learning models and fine-tuning strategies like LoRA, offer a promising alternative to the critical research challenge, especially given the diversity and large scale of satellite datasets. While deep learning-based super-resolution models have been very promising recently, their effectiveness, efficiency, and scalability across heterogeneous satellite scenes are not well studied. This work studies the performance of representative deep learning Super-Resolution frameworks, including the Enhanced Super-Resolution Generative Adversarial Network. (ESRGAN), Swin Transformer for Image Restoration (SwinIR), and latent diffusion models (LDM), under unified experimental conditions using the WorldStrat dataset. The main goal is to establish whether adaptation strategies for parameter efficiency can boost reconstruction quality while reducing computational and training costs. Toward this goal, we investigate hybrid sequential pipelines, ensemble averaging, and Low-Rank Adaptation (LoRA)–based fine-tuning. The experiments indicate that these pipelines, which use multi-model methods, achieve only marginal performance gains while incurring substantial increases in computational complexity. LoRA-Based Fine-Tuning, by contrast, has demonstrated superiority in enhancing reconstruction accuracy and quality across all model families, despite using only a small percentage of trainable parameters. LoRA-based models demonstrate superiority over multi-model methods in both efficiency and performance. The presented results confirm that LoRA is an effective and accessible technique for high-fidelity satellite-based super-resolution image synthesis. The manuscript identifies LoRA as one of the enabling technologies advancing the state of the art in Deep Learning-based Super Resolution for large-scale satellite-based image synthesis.

Computers doi: 10.3390/computers15050312

Authors: Roqia Sailh Mahmood Ismail Taha Ahmed Mohamed A. Hafez

Because textured contact lenses obscure the iris’s natural texture, they pose a serious threat to the accuracy of iris recognition systems and may make identity theft possible. Therefore, this work proposes a reliable method for textured contact lens detection that uses efficient global texture descriptors and effective feature selection with classification techniques. Run-Length and Zernike Moments are effective global texture descriptors that have been extracted from preprocessed iris images that were acquired from the IIIT-D CLI dataset. To improve classification performance, Ant Colony Optimization (ACO) was used to decrease the dimensionality of the feature vectors. Support Vector Machine (SVM) and Logistic Regression (LOG), two classifiers, have been evaluated with different descriptor pairings. According to findings from experiments, Zernike features optimized by ACO and paired with LOG produced the greatest accuracy of 98.04%, greatly surpassing previous methods. The efficacy of the presented approach for safe and dependable iris-based biometric systems is demonstrated by its exceptional results with regard to accuracy, recall, precision, and F1-score.

Computers doi: 10.3390/computers15050311

Authors: Dawei Wen Zhoujiang Peng Yuan Tian

In recent years, wildfires have occurred with increasing frequency. Pixel-level annotation of high-resolution remote sensing wildfire imagery is costly and labor-intensive. Therefore, there is an urgent need for a weakly supervised wildfire detection method that balances detection accuracy and annotation efficiency. To address the key limitations of existing weakly supervised approaches based on class activation maps (CAMs), including imprecise delineation of fire boundaries, insufficient utilization of cross-modal information, and limited capability in modeling temporal characteristics, this paper proposes a dual-branch multi-scale feature fusion framework for weakly supervised wildfire detection. The proposed framework consists of a multispectral branch and a shortwave infrared (SWIR) temporal branch, which are designed to capture the spatial structural information of fire regions and the temporal variation of thermal anomalies, respectively. Attention-guided feature fusion modules are introduced at each network stage to enable complementary integration of cross-modal information. In addition, a multi-scale CAM-weighted fusion strategy is designed to jointly enhance region localization accuracy and semantic discrimination capability. Experimental evaluations are conducted on a high-resolution wildfire dataset covering 29 regions and consisting of 2206 images. The results demonstrate that the proposed method achieves an IoU of 58.7% and an F1-score of 73.5%, outperforming the state-of-the-art methods by 4.6% and 3.2%, respectively. Ablation and comparative experiments further verify that the dual-branch architecture and feature fusion strategy significantly improve fire localization accuracy and effectively reduce the missed detection rate.

Computers doi: 10.3390/computers15050310

Authors: Farah F. Alkhalid

Problem: In the deep image steganography field, the main challenge is to achieve a balance between visual image quality, reliably recovering the message, robustness, and interpretability, especially regarding image distortion because of noise, attack, resizing, and cropping. Solution: In this paper, we propose to combine deterministic pattern-based embedding with a deep neural refinement network to achieve a strong balance between robustness, simplicity, and quality. Methodology: First of all, we embed binary messages using spatial patterns, then refine the stego image, using an encoder–decoder network and enhanced with an attention mechanism. Results: The experimental results record PSNR values between 34.9 and 37.8 dB and SSIM values above 0.99, with zero BER under no-attack, noise, and resizing conditions. Moderate degradation is observed under blur (BER ≈ 0.125), while cropping significantly affects performance (BER ≈ 0.575). Contribution: The proposed approach introduces an interpretable and stable hybrid design that effectively balances imperceptibility and robustness, while maintaining reliable message recovery in practical scenarios. The use of differentiable attacks through training enhances robustness against common distortions such as noise, blur, and resizing.

Computers doi: 10.3390/computers15050309

Authors: Ambra Korra Anduel Kuqi Indrit Enesi

Automated defect detection in industrial casting processes is important for improving product quality while reducing the cost of manual inspection. In this work, two deep convolutional neural network (CNN) architectures, InceptionV3 and InceptionResNetV2, are evaluated for the binary classification of defects in submersible pump impellers. A genetic algorithm (GA) is used to optimize key hyperparameters, including dropout rate, learning rate, and dense layer configuration, while model complexity is assessed through Pareto-based analysis. Single-run optimization results show that InceptionV3 achieves high classification accuracy (99.0%) with lower model complexity than InceptionResNetV2 (98.75%). Repeated experiments using different random seeds demonstrate relatively stable performance across runs, with InceptionV3 achieving an accuracy of 0.9913 ± 0.003 and InceptionResNetV2 achieving 0.9860 ± 0.0076. Additional experiments were conducted using random-search baselines and classification-head ablation studies (Flatten vs. Global Average Pooling). These experiments showed that optimization strategy and architectural design choices influence both predictive performance and computational complexity. The environmental impact of the training process is evaluated using CodeCarbon, with energy consumption ranging from 0.083 to 0.098 kWh and carbon emissions ranging from 2.008 to 2.401 g CO2eq for InceptionV3 and InceptionResNetV2, respectively. Overall, the results suggest that the most effective configuration depends on the evaluated architecture and experimental setting, highlighting the importance of balancing accuracy, model complexity, and computational efficiency in industrial defect detection systems.

Computers doi: 10.3390/computers15050308

Authors: Gurram Saraswathi Nagender Kumar Suryadevara

The rapid growth of Internet of Things (IoT) applications over 5G networks demands secure, low-latency data transmission while operating under strict resource constraints. However, existing studies have relied on simulations or partial implementations that fail to capture real 5G features, thus producing overly optimistic elucidations of cryptographic performance. In addition, the absence of end-to-end validation across system layers introduces an opaque flow effect, where transparency lacks across the full transmission path. To address this gap, this paper presents a fully integrated end-to-end 5G IoT security framework that introduces a modified RC4-NL (nonlinear) algorithm to enhance the security of lightweight stream ciphers while preserving computational efficiency. Environmental sensor data is encrypted on a Raspberry Pi 4B and transmitted over a commercial 5G standalone network using a Quectel FG50V module to a Multi-access Edge-Computing (MEC) server. A web-based dashboard built with FastAPI, accessed securely through an Ngrok tunnel, performs real-time decryption and visualization on 5G-connected mobile devices. This architecture eliminates the opaque flow effect and enables realistic performance evaluation, thereby avoiding the optimistic elucidations observed in simulation-based studies. This work experimentally evaluates cryptographic algorithms named Ascon, ChaCha20, AES, standard RC4, and the proposed RC4-NL under the same conditions. Experimental findings indicate that modified RC4-NL achieved an encryption time of 977 µs, a decryption time of 456 µs, and provides a lower power consumption of 0.40 watts, thus giving a proper trade-off between efficiency and enhanced security compared to standard RC4.

Computers doi: 10.3390/computers15050307

Authors: Zheli Song Yi Dong Chao Wang Xiu Zhang Aibao Sun Cuiping You Jian Mao Shuangping Liu

The objective evaluation of Daqu cross-sectional quality is challenging due to its heterogeneous structure, small features, and low contrast. This study proposes a semantic-segmentation-based framework for the automated identification and quantitative analysis of Luzhou-flavor Daqu cross-sections. Four representative architectures—including three convolutional neural network (CNN)-based models (U-Net, U-Net++, and U2-Net) and one Transformer-based model (SegFormer)—were systematically benchmarked. To address severe class imbalance and enhance model robustness, a task-specific data augmentation pipeline was implemented. With these optimized augmentation strategies, the U2-Net model demonstrated the best performance, with a peak mean Intersection over Union (mIoU) of 87.54% and a Dice score of 98.30%. Based on the predicted masks, quantitative indicators such as plaque area ratio, pizhang thickness, and fissure length were precisely extracted. The proposed framework provides an objective and scalable solution for Daqu quality inspection, offering significant practical value for industrial scenarios involving complex materials and fine-grained defect patterns.

Computers doi: 10.3390/computers15050306

Authors: Hongli Yan Xu Peng Shuang Geng Yueming Gao Junfeng Liao

Patients with chronic kidney disease (CKD) frequently experience depressive symptoms, which substantially impair their quality of life. To facilitate the early identification of high-risk individuals, this study aimed to develop a predictive model for assessing depression risk among CKD patients. This study was based on data from the China Health and Retirement Longitudinal Study (CHARLS) 2018 wave, including 1777 middle-aged and elderly participants with self-reported CKD diagnosed by a physician. Depressive symptoms were assessed using the 10-item Center for Epidemiologic Studies Depression Scale (CES-D 10). A total of 29 variables were included, covering lifestyle factors, health status, comorbidities, and sociodemographic characteristics. The Elastic Net algorithm was employed to select 11 features with the highest predictive value. Seven machine learning models, including XGBoost and support vector machine (SVM), were compared, with CHARLS 2020 data used as a temporal validation set. In the multi-model comparison, XGBoost demonstrated discrimination performance comparable to logistic regression (LR), SVM, and multilayer perceptron (MLP) (DeLong test, p > 0.05). However, considering its superior calibration performance and ability to capture nonlinear interactions, XGBoost was selected as the final model. In the validation set, the model achieved an area under the curve (AUC) of 0.8017 and an accuracy of 72.39%. SHAP analysis further revealed the nonlinear effects of predictors, with life satisfaction, sleep duration, and self-rated health showing high contributions and negative associations with depression risk, whereas limitations in activities of daily living (ADL), physical pain, and digestive system diseases were significantly associated with an increased risk of depression. Overall, the risk of depression in CKD patients is influenced by multiple dimensions, including psychological cognition, quality of life, physical function, and social environment. The predictive model developed in this study may provide a valuable reference for the early screening of high-risk populations. However, its applicability to non-CKD populations requires further validation.

Computers doi: 10.3390/computers15050305

Authors: Pius Finkel Peter Wurster

Knowledge management (KM) is increasingly becoming a critical success factor in Germany’s manufacturing industry due to demographic change, the shortage of a skilled workforce, and the growing need for flexible and resilient production systems. This study contributes empirical evidence on current KM practices in manufacturing and derives practice-oriented design implications for future LLM-supported KM systems. Two consecutive survey rounds involving six companies in Survey 1 and five companies in Survey 2 were conducted in order to identify current KM practices, recurring barriers, and design implications for large language model (LLM)-supported KM. The results show that KM is perceived as highly relevant, but is implemented only incompletely in practice. Across both datasets, central themes such as fragmented documentation practices, reliance on interpersonal transfer of tacit knowledge and uneven integration of digital KM tools recur consistently. Based on the identified practices, the paper further derives areas in which LLMs may support or augment existing KM processes, particularly with regard to semantic retrieval, contextualization, onboarding, and the preservation of tacit knowledge. The findings also highlight that successful implementation of artificial intelligence (AI)-enabled KM in manufacturing will depend on technical feasibility, trust, usability, and organizational acceptance.

Computers doi: 10.3390/computers15050304

Authors: Mazen Farid Oluwatosin Ahmed Amodu Heng Siong Lim Jamil Abedalrahim Jamil Alsayaydeh Mohammed Fadhl Abdullah Faten A. Saif

To provide reliable services in the cloud, fault tolerance is perhaps the most important consideration. The inherent sensitivity to failure hampers cloud services’ performance and reliability. As a result, fault tolerance becomes a required characteristic to maintain reliability, which is difficult to provide due to the dynamic architecture and complex inter-dependencies. To address the issues of cloud reliability, many fault-tolerant approaches have been developed in the literature. This paper presents a recent research survey that seeks to classify the various faults and intrusion tolerance architectures. Furthermore, it provides a thorough critical analysis of existing fault and intrusion tolerance, as well as combined approaches, aimed at enhancing the dependability, availability, and execution of cloud services. The report also includes a comparison of the studied systems’ framework based on various essential criteria such as cost, makespan, reliability, security, resource utilization, energy consumption, and failure ratio. This study aims to comprehensively review this subject for researchers to draw insights from existing patterns in the literature and provide deeper perspectives into some of the challenging issues and prospects. This will enhance the development of highly resilient fault-tolerant and intrusion-resistive scheduling algorithms for current and future cloud applications.

Computers doi: 10.3390/computers15050303

Authors: Yingjie Cui Xu Li Zhongxian Chen Yan Li

Fatigue seriously affects work efficiency and brings potential safety hazards, and electroencephalogram (EEG) serves as a valuable physiological indicator for fatigue monitoring, as it directly reflects underlying brain neural activity. A key characteristic in EEG fatigue research is that the feature spaces of pre-fatigue and post-fatigue EEG signals exhibit obvious spatial separation—this separation is caused by significant changes in brain electrical activity when the human body transitions from a normal awake state to a fatigue state. Existing EEG-based fatigue judgment methods mostly focus on binary classification, which fails to fully leverage the inherent spatial separation characteristic of pre-fatigue and post-fatigue feature spaces, making it difficult to achieve simple, efficient, and accurate fatigue judgment. To address this problem, this paper proposes an EEG fatigue judgment method based on feature space spatial separation and Approximate Nearest Neighbor Search (ANNS). The 16-channel pre-fatigue (Group A) and post-fatigue (Group B) EEG signals acquired from seven subjects are segmented and subjected to feature extraction, projecting the signals into a unified feature space. An ANNS index is constructed using feature vectors from both Group A and Group B, with each vector annotated by its corresponding class label. A separate test set (Group C) is utilized, and the k-nearest neighbors of each test feature vector are retrieved from the built ANNS index. The mental fatigue state is then identified via majority voting according to the class labels of the k-nearest neighbors. Experimental results demonstrate that the proposed method can effectively exploit the spatial separation between pre-fatigue and post-fatigue feature distributions, yielding an average single-subject classification accuracy of approximately 90%.

Computers doi: 10.3390/computers15050302

Authors: B. S. Premananda Mohsen Vahabi Muhammad Zohaib Seyed-Sajad Ahmadpour M. Barath K. R. Sreesha

Internet of Things (IoT) nodes continuously acquire, buffer, and transmit sensor data under strict constraints on area, latency, and energy consumption. However, conventional complementary metal–oxide–semiconductor (CMOS)-based memory-access circuits face increasing power loss, parasitic effects, interconnect complexity, and sensitivity to process variations at the nanoscale. To address these limitations, this paper proposes a quantum-dot cellular automata (QCA)-based decoder-driven static random-access memory (SRAM)-access architecture for compact and energy-efficient IoT perception-layer memory. The proposed framework integrates three main components: a majority-logic RAM cell with feedback-based storage and non-destructive readout, a compact 2 × 4 decoder with enable and auxiliary asynchronous set/reset control, and a 1 × 4 SRAM array in which the decoder is embedded to reduce routing and clocking overhead. The circuit layouts were implemented and functionally verified using QCADesigner 2.0.3, while the energy behavior was evaluated using QCADesigner-E. Simulation results confirm correct write/read (W/R) and address-selection behavior. The proposed 2 × 4 decoder achieves 86 QCA cells, 0.08 µm2 occupied area, and one clocking unit, reducing cell count, area, and clocking by 48.19%, 50.00%, and 20.00%, respectively, compared with the best selected decoder baseline. The integrated 1 × 4 SRAM array achieves 684 cells and 14 clocking units, improving timing by 30.00% compared with the closest SRAM-array baseline. These results demonstrate that the proposed QCA-based memory-access structure provides a compact and low-overhead solution for energy-constrained IoT communication systems.

Computers doi: 10.3390/computers15050301

Authors: Jarod P. Hartley W. Joseph MacInnes

Vision transformers (ViTs) have demonstrated considerable promise for classifying electrocardiogram (ECG) rhythms. However, much of the existing research is conducted in highly controlled, data-sterile settings that fail to reflect the substantial variability present in real-world ECG signals. This paper seeks to address this gap by examining how signal simplification, data quantity, and task difficulty influence the performance of the SwinV2 ViT model in ECG rhythm classification. Through systematic analysis, we highlight that classifying highly abstracted signals yields only a limited impact on model performance, with all models achieving over 95% accuracy, while the amount of training data plays a crucial role with an almost 15% accuracy difference between the models trained on the most data and the least data. Finally, our analysis shows the model’s ability to effectively adapt to an increased class count, which is essential due to the varying nature of ECG diagnosis. In summary, these results highlight the importance of carefully balancing data clarity, dataset size, and diagnostic variety when designing ECG classification systems. Achieving this balance is crucial for building reliable and scalable AI solutions for cardiac assessment.

Computers doi: 10.3390/computers15050300

Authors: Simon Leistikow Thomas Miro Adrian Kummerländer Ali Nahardani Katja Grün Marcus Franz Verena Hoerr Mathias J. Krause Lars Linsen

Hemodynamic analysis of blood flow is critical for diagnosing cardiovascular diseases and investigating cardiovascular parameters, such as aneurysms and wall shear stress. For subject-specific analyses, the anatomy and blood flow of the subject can be captured non-invasively using structural and 4D Magnetic Resonance Imaging (MRI), respectively. Computational fluid dynamics (CFD), on the other hand, can be used to generate blood flow simulations. To generate and analyze subject-specific blood flow simulations, MRI and CFD have to be brought together. We present an interactive, customizable, and user-oriented visual analysis tool that integrates measured data and CFD simulations. Thus, our open-source tool supports both medical and numerical analysis workflows. It enables the creation of simulation ensembles with a high variety of parameters. Furthermore, it allows for visual and analytical examination of simulations and measurements through 2D embeddings. To demonstrate the effectiveness of our tool, we applied it to three real-world use cases, showcasing its ability to configure simulation ensembles and analyze blood flow. We evaluated our example cases together with MRI and CFD experts. By combining the strengths of both CFD and MRI, our tool provides a comprehensive understanding of hemodynamic parameters, facilitating accurate analysis of hemodynamic biomarkers.

Computers doi: 10.3390/computers15050299

Authors: Stefanie Amiruzzaman Raga Mouni Batchu Md Amiruzzaman Linh Ngo M. Ali Akber Dewan

Digital learning tools for American Sign Language (ASL) often lack the interactive depth necessary to engage learners effectively. This paper introduces a novel, browser-based word search game designed to facilitate ASL vocabulary familiarization through gamified interaction. The system employs a two-tier architecture consisting of a React-based frontend and a Flask-based backend. At its core, the application integrates a lightweight, skeleton-based Isolated Sign Language Recognition (ISLR) model, utilizing a Stacked Transformer-based Spatial-Temporal Attention Network to enable real-time webcam-based word entry during the configuration phase. This model, trained on the WLASL-100 dataset, achieves a Top-5 test accuracy of 88.48% with an average model inference latency of 141 ms, enabling real-time webcam input without proprietary hardware. Furthermore, we implement a constraint-satisfaction puzzle generation algorithm that achieves a 100% success rate in creating interlocked, multi-directional grids. Our results demonstrate that merging computer vision with pedagogical game mechanics provides an accessible, high-performance tool for the Deaf and Hard-of-Hearing (DHH) community, bridging the gap between static instruction and active linguistic practice.

Computers doi: 10.3390/computers15050298

Authors: Xuanfei Zhou Yinxuan Huang Sining Han Jiangyao Bai

Large-scale symbolic melody datasets are essential for data-driven music information retrieval and generation, yet traditional-style Chinese melodies remain scattered across heterogeneous score formats and image sources. Existing extraction pipelines typically focus on single modalities—either MIDI archives or standard staff notation—and lack unified handling for numbered musical notation (Jianpu) and automated quality assurance. We propose the Multi-Source Melody Pipeline (MSMP), a systems-integration prototype whose front-end admits MIDI, MusicXML, Jianpu images, and staff images, and whose back-end converges on a standardized event-level representation; the present case study exercises the image branch—in particular the Jianpu branch, through a Gemini-2.5-flash vision language model—and treats the MIDI/MusicXML ingestion paths as architectural slots that are wired in but not experimentally validated in this submission. The system employs notation-aware routing to direct score images to appropriate backends (a VLM for Jianpu and rule-based OMR for staff) and enforces a structural validity gate (schema conformance plus at least one melodic track with at least one musical event) on every candidate segment. Validation on a 292-page representative prototype cohort yielded an 80.1% structural-acceptance rate—explicitly not a transcription accuracy number—and a newly added ground-truth benchmark on 50 manually annotated Jianpu pages reports 95.8% time-signature exact accuracy, 77.1% tonal-pitch-class key accuracy, 100% tempo agreement within ±5 BPM, and, on a 10-page note-level subset, a mean first-16-note pitch F1 of 0.898 (octave-sensitive) with a Symbol Error Rate of 0.150. A companion 10-page K = 3 self-consistency audit indicates that metadata errors are systematic rather than stochastic. This work, therefore, contributes a reproducible integration architecture and a quantitative baseline on the Jianpu branch, rather than a new OMR algorithm, a new dataset release, or a fully benchmarked multi-format corpus; ongoing work addresses out-of-distribution classifier evaluation, comparison against dedicated Jianpu OMR baselines, and release of a copyright-cleared corpus.

Computers doi: 10.3390/computers15050297

Authors: Bayangali Abdygalym Saule Tazhibayeva Madina Sambetbayeva Aigerim Yerimbetova Roman Taberkhan Manzura Abjalova Aidos Sabdenov Elmira Daiyrbayeva

Integrating structured linguistic resources into deep learning architectures represents a key challenge in domain-oriented NLP. This study proposes a framework for incorporating knowledge from a military thesaurus of the Ground Forces, structured according to the XML Zthes standard, into pre-trained transformed language models, including KazBERT, multilingual BERT, and XLM-RoBERTA. The approach addresses two interrelated tasks in specialized terminology processing: concept linking and semantic search. Unlike existing knowledge-injection methods designed primarily for general-domain applications, this framework formalizes the mapping of Zthes elements, such as Term, Broader term, Narrower term, Related term, ScopeNote, Language, and Source into structured textual representations that can be directly processed by transformer architectures. Fine-tuning is conducted on a dataset of 18,400 training instances automatically generated from the thesaurus, including synonym pairs, hierarchical relations (hyperonymy and hyponymy), associative links, and definitional descriptions. Experimental evaluation demonstrated that thesaurus-enriched models outperform baseline architectures across all major metrics. XLM-RoBERTA model achieves F1 = 0.84 and Top-5 accuracy = 0.94 in the concept linking task, representing a five-point improvement over the baseline. The model reaches Macro-F1 = 0.84 across four relation types. Results obtained on a specialized test set derived from terminology databases of Kazakhstan’s Armed Forces confirm robust cross-lingual generalization across Kazakh, Russian and English military discourse.

Computers doi: 10.3390/computers15050296

Authors: Stefano Palazzo Esra Hazar Arife Uslu Gokceoglu Giovanni Zambetta Roberto Caldelli Claudio Loconsole

A clustering analysis was conducted to identify distinct patient subgroups with White Blood Cells (WBC) count alongside Age and Total Immunoglobulin E (IgE) biomarkers. All data were obtained from a coordinated primary care network operating in Apulia (Southern Italy). We analyzed 300 patient records, performed preprocessing and exploratory data analysis, and then applied unsupervised clustering directly to the standardized three-variable feature space (Age, WBC, and Total IgE), followed by supervised validation steps. Several algorithms were applied for clustering. Among the evaluated methods, K-means and Spectral Clustering showed the most favorable internal validation profiles, based on Silhouette Score (SS), Calinski–Harabasz Index (CH), and Davies–Bouldin Index (DB). K-means achieved the best scores (SS = 0.406, CH = 190.00, DB = 0.900), closely followed by Spectral Clustering (SS = 0.398, CH = 182.57, DB = 0.936), outperforming Agglomerative Clustering (SS = 0.361, CH = 160.41, DB = 1.016) and Gaussian Mixture Models (SS = 0.233, CH = 103.89, DB = 1.289). Post-clustering ANOVA analyses indicated significant differences in WBC, age, and total IgE across the five consensus clusters. An evaluation of cluster internal separability occurred through the training of a Random Forest classifier to predict cluster membership. The results indicate internal cluster separability within the analyzed dataset, but more external verification and clinical evidence are necessary for validation. The research group established clinical descriptions along with suggested treatment plans and detected co-existing diseases to help validate model-based findings. A simplified cluster-informed clinical summary based on biomarker ranges was derived to support interpretation of the identified patient profiles. This integrated method preliminarily suggests that patient strata may be identified from routine clinical variables, while highlighting the importance of internal validation and clinical interpretability in clustering research.

Computers doi: 10.3390/computers15050295

Authors: Yan Zhang

Machine learning and statistical learning have become central to modern scientific discovery and technological innovation [...]

Computers doi: 10.3390/computers15050294

Authors: Assem Turarbek Diana Rakhimova Yeldos Adetbekov Azat Nurgali

The rapid growth of heterogeneous web sources has created significant challenges for the automated extraction and structuring of critical domain-specific information, particularly in real-time seismic monitoring scenarios. Despite the existence of official governmental reporting systems, relevant earthquake-related data are often distributed across diverse online platforms with highly variable and dynamically evolving HTML (HyperText Markup Language) structures, leading to incomplete, delayed, or inconsistent information retrieval. Existing rule-based and semi-automated approaches lack scalability and robustness under such conditions. To address this gap, this study proposes a self-adaptive framework based on large language models (LLMs) for the automated extraction and structuring of earthquake-related web content. The proposed approach integrates transformer-based schema generation, repository-guided schema matching, and an iterative refinement mechanism, enabling the system to dynamically adapt to heterogeneous document structures. A formal utility-based decision mechanism is introduced to optimize schema selection and reuse, while embedding-based similarity modeling facilitates efficient transfer of extraction patterns across structurally related webpages. The experimental evaluation was conducted on a heterogeneous benchmark dataset comprising multiple web domains with diverse structural characteristics. The results demonstrate that the proposed framework achieves a success rate of 85% across all evaluated models, with the best-performing configuration reaching an extraction accuracy of 96.5% and a final composite score of 84.26. Additional analysis reveals significant improvements in extraction completeness, reduction in false positives and false negatives, and effective reuse of a compact set of robust schemas. Error analysis indicates that the primary challenges are associated with noisy HTML structures and incorrect DOM (Document Object Model) element selection, rather than deficiencies in textual content. The findings confirm that combining lightweight transformer models with adaptive memory and schema reuse mechanisms enables the development of scalable, robust, and high-performance web extraction systems. The proposed approach is particularly suitable for real-time information retrieval in safety-critical domains, where timely and accurate data aggregation from heterogeneous sources is essential.

Computers doi: 10.3390/computers15050293

Authors: Leiyue Yao Zhiqiang Dai Keyun Xiong

Squats are essential for assessing lower limb strength. However, performing them incorrectly without professional guidance often leads to sports injuries. Currently, most detection methods rely heavily on deep neural networks and massive datasets. This approach brings several downsides. It involves high data labeling costs and heavy computing demands. It is also difficult to achieve low-latency feedback on mobile devices. Furthermore, these models often lack robustness when dealing with individual body differences. To tackle these issues, we propose a new real-time squat detection method. Our approach is built on prior rules and statistical models. Here is how it works. First, we use MediaPipe to track the body’s skeleton joints in real-time from video feeds, calculating the hip and knee angles frame by frame. Next, we build a hip-knee coordination model using linear regression. This step helps us measure how these joints move together dynamically. Finally, we verify the squat depth using a geometry-based tolerance mechanism. This feature accounts for measurement noise and natural body variations, allowing us to accurately judge if the overall posture is standard. We tested our approach on three different squat styles. The results show that our method catches improper forms quickly and efficiently in real time, achieving an accuracy of 90%.

Computers doi: 10.3390/computers15050292

Authors: Aleksandr Yurin Nikita Dorodnykh

End-user development (EUD) and model-driven engineering (MDE) are particularly valuable for building classical intelligent systems that rely on declarative knowledge bases. In these knowledge bases, the key dependencies of the domain can be described in the form of logical rules. The general-purpose modeling language used in MDE, specifically UML, enables modeling of static data structures and the dynamics of object behavior; however, it does not primarily support the modeling logical rules. In this paper, we propose a rule visual modeling language inspired by UML—Rule Visual Modeling Language (RVML)—which expands the capabilities of MDE in terms of using domain-specific visual languages. This approach substantially supports end-users in constructing declarative knowledge bases. We present the formal semantics, visual syntax, and features of RVML, along with two industrial case studies. We empirically evaluate the effectiveness of RVML in development compared to other graphic notations used for modeling logical rules. Our evaluation demonstrates that RVML provides superior expressiveness and better preservation of semantic integrity.

Computers doi: 10.3390/computers15050291

Authors: Shweta Neha Gupta Meenakshi Gupta Massimo Donelli Yogita Arora Achin Jain

Fetal health assessment is a crucial aspect of prenatal care, aimed at the early detection of potential complications to ensure optimal outcomes for both mother and child. Traditional methods, such as the visual analysis of cardiotocography (CTG) data by healthcare professionals, are valuable but often subjective and time-consuming. This work investigates the application of machine learning techniques, with a focus on ensemble learning, to enhance the accuracy and efficiency of fetal health classification based on CTG data. Genetic Algorithm (GA) is employed for optimal feature selection, identifying the most discriminative subset of CTG attributes to improve model performance and reduce computational complexity. We employ a combination of advanced machine learning models, including AdaBoost, Gaussian Naïve Bayes, Decision Tree, k-nearest neighbors (KNN), and Logistic Regression. The top two models were selected based on comprehensive performance metrics using the TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) method. These models were then integrated through ensemble learning approaches, such as stacking, Particle Swarm Optimization (PSO) weighted averaging, and soft voting, to improve prediction reliability. Our proposed stacking ensemble model achieves a remarkable accuracy of 97.9%, demonstrating its potential as a robust, data-driven tool for fetal health monitoring and the early identification of at-risk pregnancies. The results indicate that machine learning can effectively complement traditional fetal health assessment methods by providing an objective framework to support clinical decision-making.

Computers doi: 10.3390/computers15050290

Authors: Duaa J. Khalaf Arwa A. Moosa Tayseer S. Atia

The integration of post-quantum digital signature (PQDS) algorithms into coherent wavelength-division multiplexing (WDM) optical networks introduces a non-negligible cryptographic overhead that fundamentally alters physical-layer performance characteristics. Unlike conventional studies that treat security and transmission independently, this work provides a cross-layer evaluation of PQDS-induced payload expansion and its direct impact on coherent optical system behavior under realistic, DSP-aligned conditions. A structured and reproducible evaluation framework is proposed to systematically analyze this interaction across multiple transmission scenarios, ranging from a single-channel QPSK baseline to a 16-channel WDM system employing both QPSK and 16QAM modulation formats. Key system parameters—including launch power, local oscillator power, bit rate, and fiber length—are jointly optimized, while performance is rigorously assessed in terms of bit error rate (BER), Q-factor, and maximum transmission reach. The results demonstrate a clear performance degradation trend driven by both spectral efficiency scaling and cryptographic payload expansion. The single-channel QPSK system achieves a maximum reach of 203 km, which decreases to 194 km in the 16-channel WDM QPSK configuration due to inter-channel interference and nonlinear effects. In contrast, the 16-channel WDM 16QAM system exhibits a significantly reduced reach of 103 km, reflecting its heightened sensitivity to noise, chromatic dispersion, and fiber nonlinearities. Furthermore, increased payload size associated with PQDS schemes is shown to exacerbate transmission impairments by extending frame duration and intensifying inter-channel interactions. These findings identify PQDS-induced overhead as a critical system-level constraint that directly governs transmission efficiency, scalability, and performance limits. The study highlights the necessity of cross-layer co-design strategies, where cryptographic mechanisms and physical-layer parameters are jointly optimized to enable efficient, reliable, and quantum-safe coherent optical communication systems.

Computers doi: 10.3390/computers15050289

Authors: Islem Sahraoui Kinam Kim Lu Gao Zia Ud Din Ahmed Senouci

This study investigates the educational potential of Large Language Models (LLMs) for automating rule-checking tasks in Building Information Modeling (BIM) instruction. A quasi-experimental classroom implementation was conducted over two consecutive semesters with 55 graduate students in a Construction Management program. In Fall 2024, students were taught manual rule-checking techniques, whereas in Spring 2025, students received additional instruction in LLM-based prompting and Python code generation for automated compliance checking. A mixed-methods evaluation was conducted using surveys, NASA Task Load Index ratings, assignment-based learning outcomes, and structured interviews. Compared with the manual-only cohort, the LLM-assisted cohort reported significantly lower mental, temporal, and frustration demands, as well as higher perceived time efficiency and overall effectiveness. The LLM-assisted group also achieved significantly higher performance in violation detection and method accuracy, although no significant differences were observed in code interpretation or reflective analysis. Qualitative findings further revealed both the efficiency benefits of AI-assisted automation and persistent challenges related to prompt refinement, debugging, and output validation. These findings suggest that LLMs can enhance BIM instruction when paired with structured pedagogical scaffolding to support critical oversight and novice learners.

Computers doi: 10.3390/computers15050288

Authors: Fangfang Shan Zhuo Chen Yifan Mao Yuhang Liu Lulu Fan Yanlong Lu

Federated learning, a decentralized machine learning framework, allows multiple participants to jointly train models while keeping their raw data local and unshared. Nevertheless, during the exchange of model updates, the communicated information can still introduce privacy vulnerabilities and potentially result in the exposure of user data. Over the past few years, differential privacy methods have been broadly incorporated into federated learning frameworks to strengthen the protection of sensitive data. Nevertheless, the noise required to satisfy differential privacy guarantees often causes significant degradation in model performance. Prior studies have typically employed a fixed noise-injection strategy following gradient clipping. Although such methods provide privacy protection, they overlook the varying importance of different gradient dimensions, resulting in noise being injected into unimportant or redundant parameters, thereby causing unnecessary performance loss. To address these limitations, we propose an adaptive differentially private federated learning scheme with importance-based gradient updates (AIGU-DPFL). Specifically, we focus on coordinates with high information content and introduce an adaptive noise injection mechanism, which perturbs gradient updates to satisfy differential privacy guarantees while dynamically controlling noise intensity, thereby achieving sparse and noise-effective gradient updates. AIGU-DPFL markedly enhances the training effectiveness of federated learning models. Comprehensive evaluations conducted on real-world datasets indicate that the proposed method achieves superior performance compared to existing differentially private federated learning techniques.

Computers doi: 10.3390/computers15050287

Authors: Lara Marie Reimer Leonard Pries Florian Schweizer Leon Nissen Stephan M. Jonas

Speech alterations are early markers of neurodegenerative diseases. Transformer-based speech models such as Whisper have advanced automated speech assessment, but most systems rely on cloud-based computation, raising privacy concerns. On-device processing could offer a scalable and privacy-preserving alternative. This research’s objective was to evaluate whether a fully on-device speech analysis pipeline can achieve competitive accuracy in detecting Alzheimer’s disease and to quantify the contributions of acoustic, linguistic, and embedding features. Therefore, we developed an iOS application running all components, including acoustic analysis, two transformer-based speech-to-text modules (WhisperBase and quantized CrisperWhisper), linguistic feature extraction, and embedding generation, directly on the device. Using the ADReSS Challenge 2020 dataset (N = 156), we trained classical machine-learning classifiers across 20 configurations and evaluated them via a stratified 10-fold cross-validation. Area under the receiver operating curve (AUC), accuracy, precision, recall, and F1 scores were used as performance metrics. An ablation study examined the relevance of each feature group. The best-performing setup (Random Forest with CrisperWhisper transcription and Apple embeddings) achieved an accuracy of 85.4% and an AUC of 0.85. Performance was 5–7% below benchmark models relying on manual transcripts or server-based processing. Embedding features provided the strongest individual contribution, but the highest accuracy required combining acoustic, linguistic, and embedding information. A fully on-device pipeline for Alzheimer’s disease detection from speech is feasible and achieves competitive accuracy while maintaining strict data privacy. These findings highlight the potential of on-device transformer architectures for scalable, privacy-preserving digital screening. Future work should validate the approach in larger and more diverse cohorts.

Computers doi: 10.3390/computers15050286

Authors: Degang Zhu Shiqi Wei Xinming Zhang

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 μ 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.

Computers doi: 10.3390/computers15050285

Authors: Nouf Alkhater

The increasing complexity of modern network infrastructures has intensified the need for reliable and efficient intrusion detection systems. While advanced deep learning approaches have demonstrated strong performance, their high computational cost and limited interpretability restrict their practical deployment in real-time environments. This study presents a systematic empirical evaluation of four supervised machine learning models—Decision Tree, Random Forest, Support Vector Machine (SVM), and XGBoost—for network anomaly detection using the UNSW-NB15 dataset. To ensure methodological rigor, a structured preprocessing pipeline and a five-fold stratified cross-validation framework were employed. Model performance was assessed using multiple evaluation metrics, including accuracy, precision, recall, F1-score, and area under the ROC curve (AUC). In addition, a feature importance analysis was conducted to identify the most influential network traffic attributes contributing to anomaly detection. The results show that ensemble-based methods outperform individual classifiers, with XGBoost achieving the best overall performance (accuracy = 0.97, AUC = 0.98) along with high stability across validation folds. The analysis further reveals that a subset of flow-based and temporal features—such as sttl, sload, and dload—plays a critical role in distinguishing between normal and malicious traffic. This study provides a rigorous, interpretable, and reproducible benchmarking framework for supervised machine learning in network anomaly detection. The findings provide practical insights for developing efficient and scalable intrusion detection systems suitable for real-world deployment.

Computers doi: 10.3390/computers15050284

Authors: Xinyi Jiang Muying Luo Zainuddin Ibrahim Azlan Abdul Aziz Azhar Jamil

Although motion capture has become integral to contemporary animation pipelines, university teaching still asks students to learn motion largely through screen-based keyframing. To address this gap, this classroom-based comparative study evaluated one structured motion-design lesson within an immersive MoCap-supported training module. Sixty-eight undergraduates in a computer animation course completed the same task in either a Keyframe condition (n = 33) or a VR-MoCap condition (n = 35), with instructional delivery mode as the only difference. Creative performance was assessed in originality, fluency, aesthetic quality, clarity, and a composite score. MANOVA revealed a significant multivariate effect of condition (Pillai’s trace = 0.454, F(4, 63) = 13.12, p < 0.001). Relative to keyframe instruction, VR-MoCap produced significantly higher originality, fluency, clarity, and composite performance, whereas aesthetic quality did not differ significantly. Supplementary group-interview responses further indicated that students experienced the immersive condition as more engaging, more intuitive, and better suited to immediate feedback and embodied movement exploration. Immersive VR-MoCap appears most useful in the early phases of motion design and is better understood as complementing, rather than replacing, conventional keyframe training.

Computers doi: 10.3390/computers15050283

Authors: Shiren Ye Liangjing Li Zetong Zhang Haipeng Ma

Accurate detection of small intestinal lesions in wireless capsule endoscopy (WCE) images remains challenging because lesions are often small, weakly contrasted, irregular in shape, and easily confused with complex mucosal backgrounds. To address these difficulties, this study proposes YOLOv12-WCIRS, a WCE-oriented improvement of YOLOv12 that jointly enhances local feature extraction, selective multi-scale fusion, background suppression, localization sensitivity, and scale-aware optimization. The proposed framework incorporates a Weighted Convolution (WConv) module, a Contextual Selection Fusion Module (CSFM), an Information Integration Attention Fusion (IIA_Fusion) module, a Receptive Field Attention-based detection head (RFAHeadDetect), and a Scale Dynamic Loss (SD Loss). Experiments on the SEE-AI dataset show that YOLOv12-WCIRS achieves 83.4% mAP@0.5 and 61.1% mAP@0.5:0.95, improving mAP@0.5 from 76.9% to 83.4% over the direct baseline YOLOv12 while maintaining competitive efficiency. Additional analyses, including cross-dataset validation on overlapping categories in Kvasir-Capsule, normal-frame false-alarm evaluation, false-positive/false-negative breakdown, and repeated-run statistical testing, further support the robustness and practical value of the proposed framework. These results indicate that YOLOv12-WCIRS provides an effective solution for automated lesion detection in WCE images and shows promise for computer-aided capsule endoscopy analysis.

Computers doi: 10.3390/computers15050282

Authors: Jasmin Redžepagić Ana Kapulica Nikola Malešević Vedran Dakić

This study presents a systematic empirical evaluation of monolithic and distributed database architectures deployed in Kubernetes environments. As containerized and cloud-native infrastructures become increasingly prevalent, understanding the performance implications of running stateful data systems under orchestration platforms has become critical. We evaluate five widely used database systems—PostgreSQL, MySQL, MongoDB, Redis, and Cassandra—using standardized workload generation frameworks, including pgbench, sysbench, YCSB, redis-benchmark, and cassandra-stress. Controlled experiments were conducted across varying concurrency levels and workload types to measure throughput, latency, and scalability in both single-node and distributed deployments. Redis achieves a maximum throughput of 4.2 million operations per second with sub-millisecond latency. In contrast, Cassandra delivers 214,743 distributed read operations per second at ONE consistency, approaching Redis’s non-pipelined baseline throughput (257,732–262,467 ops/sec) within a Kubernetes cluster. The write throughput of Cassandra decreases by 45.2% when the consistency level is elevated to QUORUM, accompanied by an elevenfold increase in run-to-run variability (CV from 7.1% to 84.7%), indicating that the consistency level is the primary performance determinant in distributed systems. PostgreSQL experiences a 72% decrease in write throughput in Kubernetes (74,072 → 20,805 TPS). In contrast, MySQL PXC anomalously attains a 37.3% increase in write throughput in Kubernetes compared to its monolithic deployment—the sole reversal noted among the five systems. These findings underscore a critical trade-off between vertical efficiency and horizontal scalability, illustrating that hybrid database architecture can be an effective solution for contemporary cloud-native applications compared to either paradigm independently.

Computers doi: 10.3390/computers15050281

Authors: Yongmei Tan Gong Chen Yan Huang Hengzhou Ye Jincheng Tang

The Swin Transformer exhibits limitations in fine-grained land use and land cover (LULC) classification, particularly in capturing high-frequency texture details and representing low-contrast regions. To address these issues, we propose a novel network model, termed LACE-Net, which integrates local frequency-domain energy and adaptive contrast enhancement. Built upon the Swin Transformer backbone, the model introduces an innovative Local Frequency-Domain Energy-Adaptive Contrast Enhancement Multi-Scale Attention (LACE). This block consists of parallel branches for frequency-domain perception and contrast enhancement, which effectively combine texture and illumination physical priors. In addition, a texture-adaptive momentum adjustment mechanism is incorporated to refine the spatial enhancement attention weights dynamically. Consequently, LACE-Net greatly strengthens the modeling and representation of high-frequency details and complex spatial structural features. Experiments are performed on a self-constructed Guangxi regional dataset (denoted as GLC-30) and the publicly available remote sensing scene classification benchmark dataset NWPU-RESISC45. The results show that LACE-Net achieves a Top-1 accuracy (Top-1 Acc) of 96.48% and a macro-averaged F1 score (mF1) of 93.13%. These results outperform current mainstream vision models, particularly in mitigating the spectral confusion issue of “same spectrum, different objects.” The model exhibits superior fine-grained classification performance and robust generalization across datasets.

Computers doi: 10.3390/computers15050280

Authors: Junbo Xiang Tiejun Wang

Efficiently and accurately converting floating-point numbers to decimal strings remains a fundamental challenge in numerical computation, data serialization, and human–computer interaction. While modern algorithms such as Ryū, Dragonbox, and Schubfach rigorously satisfy the Steele–White criteria for correctness and minimal output length, their performance is frequently constrained by branch mispredictions, high-precision multiplication overhead, and suboptimal utilization of instruction-level parallelism. This paper introduces xjb, a novel floating-point–string conversion algorithm derived from Schubfach that systematically overcomes these bottlenecks. By restructuring the core computation to reduce instruction dependencies, adopting branchless decision logic, and exploiting SIMD instruction sets for decimal-to-ASCII formatting, xjb delivers state-of-the-art throughput across diverse hardware platforms. The algorithm requires only a single 64-by-128-bit multiplication for IEEE 754 binary64 conversions and a single 64-by-64-bit multiplication for binary32, drastically decreasing arithmetic complexity. Extensive benchmarking on AMD R7-7840H and Apple M1/M5 processors demonstrates that xjb consistently outperforms leading contemporary implementations. Notably, on the Apple M5, xjb achieves speedups of approximately 20% and 136% for binary64 and binary32 conversions, respectively, when compared to the highly optimized zmij library. The algorithm is fully compliant with the Steele–White principle; exhaustive validation over the entire binary32 space and extensive random testing across the binary64 range confirm both its theoretical soundness and practical robustness.

Computers doi: 10.3390/computers15050279

Authors: H M Quamran Hasan Housam Khalifa Bashier Babiker Mi-Young Kim Randy Goebel

Large Language Models (LLMs) used for clinical decision support must not only make accurate predictions but also generate rationales that are consistent with, and sufficient for, those predictions. Building on Reason2Decide, a two-stage rationale-driven multi-task framework, we propose Reason2Decide-C (R2D-C, where C denotes cycle consistency), which augments Reason2Decide’s stage 2 training with confidence-adaptive scheduled sampling and cycle-consistent rationale-to-label training. In stage 1, we pretrain our model on rationale generation. In stage 2, we jointlytrain on label prediction and rationale generation, gradually replacing gold labels with model-predicted labels based on confidence. Simultaneously, we feed the rationale logits back into the model to recover the label, thus enforcing explanation sufficiency. We evaluate R2D-C on one proprietary triage dataset, as well as public biomedical QA and reasoning datasets. Across model sizes, R2D-C substantially improves rationale–prediction consistency (where stage 1 and stage 2 predictions agree) and sufficiency (where the rationale alone recovers the ground-truth label) over other baselines while matching or modestly improving predictive performance (F1); in several settings R2D-C surpasses 40× larger foundation models. Ablations confirm that the full combination is optimal, maximizing alignment and LLM-as-a-Judge rationale quality. These results demonstrate that confidence-adaptive scheduled sampling and cycle-consistent rationale-to-label training substantially enhance explanation alignment without sacrificing accuracy.

Computers doi: 10.3390/computers15050278

Authors: Andrés Eduardo Fuentes-Cortázar José Rafael Rojano-Cáceres

Mobile devices are widely used, yet accessibility for people with disabilities remains a critical challenge. Deaf users who rely primarily on sign language (SL) frequently encounter barriers when interacting with applications not designed for their communication needs. This study proposes a design guide for developing mobile applications tailored to sign language users. The guide was developed through the active participation of three groups: Deaf individuals, usability and user experience (UX) experts, and mobile application developers. Based on their contributions, thirteen design guidelines were defined, addressing sign language integration, visual feedback, navigation, content presentation, and interface design. The guidelines were validated through usability and UX evaluations conducted with the three participant groups. A mobile application was subsequently developed following the proposed guidelines to assess their practical applicability. The evaluation results indicate that the guide effectively supports the development of more accessible and usable mobile applications for Deaf users. Incorporating sign language-centered design principles significantly improves usability and user experience for individuals with hearing disabilities, contributing to more inclusive mobile application development.

Computers doi: 10.3390/computers15050277

Authors: Lev Sukherman Yetunde Longe-Folajimi Marina Konkol

This paper investigates whether structured collaboration between multiple large language models (LLMs), each assigned a distinct cognitive role grounded in psychological theory, produces benefits beyond simple answer aggregation. We propose the Parallel Synthesis architecture, in which three cognitively specialized roles Analyzer (hierarchical decomposition), Creative (divergent thinking), and Critic (critical evaluation) process each task independently and in parallel, and a Synthesizer integrates their outputs into a final response. To evaluate collaborative reasoning, we introduce the Emergent Reasoning Score (ERS), a composite metric that separates perspective integration (Synthesis Effectiveness) from novel concept generation (Emergent Value). Experiments on Experiments on the AI2 Reasoning Challenge (ARC-Challenge) (1172 questions) and and the Massive Multitask Language Understanding benchmark (MMLU) (1531 questions) show two consistent findings. First, the architecture achieves high Synthesis Effectiveness (SE=0.711–0.744), indicating reliable integration of all three cognitive perspectives. Second, Emergent Value remains low (EV=0.096–0.112), indicating that synthesis primarily recombines existing concepts rather than generating substantial novel content. A Majority Voting baseline achieves comparable or slightly higher answer accuracy than the Synthesizer on both benchmarks, showing that the architecture’s main contribution lies not in answer selection but in producing integrated reasoning traces that draw on multiple perspectives. These findings suggest that the practical value of cognitively grounded multi-agent architectures lies in reliable perspective integration, while ERS provides a reusable framework for distinguishing integration from genuinely novel reasoning in multi-agent LLM systems. The empirical results reported here constitute a pilot validation of the proposed framework on closed-form benchmarks, intended to establish a proof of concept and motivate larger-scale evaluation.

Computers doi: 10.3390/computers15050276

Authors: Stavroula Chatzinikolaou Giannis Vassiliou Efstratia Vasileiou Sotirios Batsakis Nikos Papadakis

Continuous customer activity generated by modern digital platforms drives the evolution of behavioral segments over time. Traditional customer segmentation methods typically rely on periodic batch analysis of historical data, producing static snapshots that may quickly become outdated and fail to capture emerging behavioral patterns. This paper presents a monitoring-oriented framework for detecting customer segment evolution and generating timely notifications about meaningful structural changes in the customer population. The proposed system continuously ingests user activity events, incrementally updates customer profiles, and periodically recomputes behavioral segments using fixed-k KMeans clustering over standardized recency, frequency, and monetary (RFM) features. To improve robustness and interpretability, the framework incorporates adaptive event scoring, stability-aware segment validation, drift-aware centroid matching, and persistence-based filtering of transient changes. These mechanisms reduce noisy alerts caused by repeated clustering updates while preserving meaningful signals about evolving customer behavior. The framework is evaluated on the Online Retail II and Instacart datasets under streaming simulation conditions. Experimental results show that the proposed approach maintains stable clustering structures, identifies persistent segment changes, and uncovers economically meaningful customer groups. Compared with static segmentation and periodic clustering baselines, the framework improves clustering quality while enabling continuous monitoring of segment evolution. Overall, the results suggest that adaptive monitoring can extend traditional customer segmentation into a practical continuous analytics process for moderate-scale dynamic environments.

Computers doi: 10.3390/computers15050275

Authors: Rao Mikkilineni William Patrick Kelly

Contemporary enterprise IT operations are largely implemented on Shannon–Turing computing models in which programs execute read–compute–write cycles over data structures, while governance—fault handling, configuration control, auditability, continuity, and accounting—is applied externally through infrastructure platforms, observability stacks, and human operational processes. This separation scales analytical throughput but accumulates what we term coherence debt: locally expedient operational commitments whose provenance and revisability degrade over time until exposed by failures, security incidents, regulatory demands, or architectural transitions. This paper examines the evolution of operational computing models that integrate com-pupation with regulation at two distinct levels. First, Distributed Intelligent Managed Elements (DIME) extend the classical Turing cycle toward a supervised execution loop—read–check-with-oracle–compute–write—by incorporating signaling overlays and FCAPS (Fault, Configuration, Accounting, Performance, and Security) supervision into computation in progress. Second, the Autopoietic Management and Orchestration System (AMOS), grounded in the General Theory of Information, the Burgin–Mikkilineni Thesis, and Deutsch’s epistemic framework, fully decouples process executors from governance by treating any Turing-equivalent engine as a replaceable execution substrate while elevating knowledge structures—encoded as local and global Digital Genomes—to first-class operational state within a governed knowledge network. Using a distributed microservice transaction testbed, we demonstrate how this approach operationalizes topology-as-data, a capability-oriented control plane, decoupled application-layer FCAPS independent of infrastructure management, and policy-selectable consistency/availability semantics. Our results show that the principal benefit of AMOS is not circumventing theoretical constraints such as the Consistency, Availability, and Partition tolerance (CAP) theorem, but governing their trade-offs as explicit, auditable commitments with defined convergence pathways and controlled return to a coherent system state, thereby reducing coherence debt and improving operational reliability in distributed AI-enabled enterprise systems.

Computers doi: 10.3390/computers15050274

Authors: Vanesa Landero-Nájera Joaquín Pérez-Ortega Laura Cruz-Reyes Claudia Guadalupe Gómez-Santillán Nelva N. Almanza-Ortega Carlos Rodríguez-Orta Carlos Andrés Collazos-Morales

The logical structure of approximation algorithms has been identified by the scientific community in four principal parts: tuning parameters, generating initial solutions, generating neighbor solutions, and stopping algorithm execution. A review of the literature specifically for the algorithms Threshold Accepting (TA) and Tabu Search (TS) indicates that, in most cases, choices are performed on one or several of these logical parts, often implicitly guided by expert knowledge for improving algorithm performance. However, these design choices, particularly in the selection of initialization and neighborhood strategies, are rarely analyzed in a systematic and reproducible manner. A formal experimental framework is presented to systematically analyze logical structure design choices, which are typically based on empirical expertise, by isolating and evaluating the combined effects of methodologies in the logical parts of initialization and neighborhood under controlled conditions of TA and TS algorithms in solving the one-dimensional Bin Packing Problem (BPP). A total of 324 benchmark instances were used to assess multiple algorithmic variants. Performance was evaluated in terms of solution quality and computational effort, supported by graphical analysis and statistical methods, including Wilcoxon signed-rank tests, effect size measures, bootstrap-based confidence intervals, and linear regression. The experimental results consistently show that the simpler internal logical structure of TA and TS algorithms, specifically with a probability-guided initialization combined with a single neighborhood operator, can achieve a better balance between solution quality and computational effort compared to more complex alternatives in general instances of BPP.

Computers doi: 10.3390/computers15050273

Authors: Xianlang Hu Guangsheng Feng Xinling Huang Xiangying Kong Hongwu Lv

System logs have been critical for analyzing the operational status and abnormal behavior of highly distributed and heterogeneous edge computing nodes. In edge environments, logs exhibit cross-event and cross-field structural interactions, making it difficult to uncover potential anomaly patterns from isolated events. Moreover, sparse annotations and varying log formats limit the effectiveness of existing methods. To address these challenges, we propose a graph neural network (GNN) anomaly detection framework with prompt learning. It leverages few-shot prompt learning to automatically extract key fields and constructs a weighted directed graph that jointly models semantic embeddings and temporal dependencies, fully representing the structural interactions and semantic associations across events and fields. Furthermore, the framework performs graph-level anomaly detection by jointly optimizing graph representation learning and classification objective within an enhanced one-class directed graph convolutional network, enabling effective identification of global structural anomaly patterns in log graphs. Experimental results demonstrate that the proposed method achieves an average F1-score of 93.3%, surpassing the current state-of-the-art (SOTA) methods by 6.93%.

Computers doi: 10.3390/computers15050271

Authors: Andrew N. Shafik Mohamed H. Khafagy Alber S. Aziz Shereen A. Hussein

Instance segmentation in heterogeneous waste scenes remains challenging due to object variability, deformable shapes, partial occlusion, and large appearance differences across packaging types. This study presents a YOLOv11-based deep learning framework for mixed plastic waste instance segmentation, developed to connect visual perception with reliable material quantification. The framework integrates curated instance-level annotations, strict split isolation, multi-stage optimization, training strategy ablation, and seed-robustness analysis to support reproducible model selection. Experimental results on a held-out test set show that the optimized model achieves a mask mAP@50:95 of 0.9337, indicating strong segmentation performance under heterogeneous waste-scene conditions. To extend the analysis beyond standard vision metrics, the framework incorporates a physics-informed mask-to-mass module that converts predicted masks into class-specific mass estimates using geometric calibration and material priors. Applied to a representative stream of 1253 detected objects, the system estimated a total plastic mass of 15.48 ± 1.08 kg, corresponding to a theoretical H2 potential of 0.41 ± 0.04 kg and a greenhouse-gas avoidance of 34.57 ± 4.15 kg CO2e. Overall, the proposed framework extends waste-scene understanding beyond vision-level assessment toward physically grounded, data-driven decision support for smart material recovery systems.

Computers doi: 10.3390/computers15050272

Authors: Raja Waseem Anwar Flavio Pastore Tariq Abdullah

Despite its continued growth, cloud computing remains susceptible to significant security challenges, as shared virtualised environments pose threats at multiple levels. These vulnerabilities are caused by a lack of security coverage in the responsibility model between the provider and the tenant. In this work, we propose the multi-layered architecture VIRTUOSO (VIRTual Unified Operation Security Optimiser) to cover these security gaps through advanced automation and ML. VIRTUOSO has four layers. The Input Layer extracts key risk components from collected telemetry data. The Deep Automation Security Layer provides automated actions and continuous monitoring of security defences. Its counterpart, the Intelligent Security Layer, predicts threats using anomaly detection. The last layer, the Output Layer, returns an aggregated risk summary. The datasets we used were chosen for their relevance: the UNSW-NB15 dataset, a subset of the web-attack classification from CSE-CIC-IDS2018, and a sample of anonymised log events from AWS CloudTrail. Our ensemble classifiers achieve a best accuracy of 95.08% ± 0.13% on UNSW-NB15 (RF), with statistically significant differences among models confirmed by the Friedman test (p < 0.004) and Nemenyi post hoc analysis, and 99.25% ± 0.52% on web-attack (CatBoost), where ensemble differences are not statistically significant (p = 0.093), consistent with the high separability of this dataset. The training-test gap and DNN curves show no overfitting, whereas our adversarial tests show a maximum accuracy loss of 8.1% at ε = 0.02. With these promising results, we can assert that, pending verification in an actual cloud environment and potential integration with FL, our ensemble classifier model appears to be a good real-world prototype.

Computers doi: 10.3390/computers15050270

Authors: Julio Gómez López Deian Orlando Petrovics Tabacu Nicolás Padilla Soriano Alfredo Alcayde García

Network Intrusion Detection Systems (NIDSs) are critical to ensuring the resilience of modern digital infrastructures. Although traditionally deployed in large-scale corporate environments, the expanding threat landscape requires the integration of robust security measures into Small Office/Home Office (SOHO) and Edge-of-Things (EoT) networks. However, these environments often face significant constraints in terms of specialized hardware and technical expertise. This article presents R-Snort, an open-source NIDS based on Snort 3, optimized for low-cost Raspberry Pi 5 hardware. Its multi-agent architecture enables distributed deployment with centralized traffic analysis and cross-agent attack correlation, while an intuitive web interface simplifies alert visualization and system management for non-expert administrators. Its main contributions are: (1) a performance-optimized NIDS agent achieving 1 Gbps throughput; (2) a distributed multi-agent architecture enabling centralized event correlation and detection of multi-vector attacks; and (3) an IaC-based automated deployment framework with an intuitive web interface, democratizing professional-grade security for SOHO and EoT environments.

Computers doi: 10.3390/computers15050269

Authors: Min Yang Makoto Oka Hirohiko Mori

In this study, we developed a chatbot-based system for detecting early signs of depression and verified its effectiveness through experimental evaluations and user surveys. Emphasizing that it does not rely on medical checklists, the system is designed to automatically extract three linguistic features associated with depression—frequent use of first-person pronouns, pessimistic expressions, and obsessive-compulsive writing styles—from natural user conversations. Multiple models were constructed for these features, and an ensemble layer integrates their outputs for a comprehensive judgment. The implemented system analyzes input sentences obtained through chat, extracts the three categories of features, calculates a final score through an ensemble layer, and visualizes potential signs of depression based on the total score. We performed an evaluation experiment with 20 participants. In the test data evaluation, the system demonstrated over 76% accuracy in each of the three classification categories: first-person usage, pessimistic tendency, and obsessive-compulsive tendency.

Computers doi: 10.3390/computers15050268

Authors: Mussa Turdalyuly Ainur Tursynkhan Aigerim Yerimbetova Tolganay Turdalykyzy Bakzhan Sakenov Nurzhan Mukazhanov Nazerke Baisholan

Learning meaningful representations of scientific documents is essential for information retrieval, knowledge discovery, and recommendation systems. Traditional methods such as TF-IDF rely on lexical matching and fail to capture deeper semantic relationships, while transformer-based approaches typically depend on limited supervision signals. In this work, we propose a Triple-Source automatic supervision framework for learning document embeddings from scientific corpora. The model integrates three types of supervision–title–abstract pairs, same-category document pairs, and document-level semantic relationships—within a unified contrastive learning framework based on a multilingual XLM-RoBERTa encoder. Unlike prior approaches that rely on citation graphs or manual annotations, our method enables citation-free and annotation-free representation learning using only lightweight metadata. Experiments on a publicly available arXiv dataset consisting of 98,649 documents demonstrate improved semantic retrieval performance, achieving Recall@1 = 0.6181 for same-category retrieval and outperforming both TF-IDF and single-source transformer baselines. The learned embeddings also exhibit improved clustering of scientific domains, indicating more structured semantic representations.

Computers doi: 10.3390/computers15050267

Authors: Aida Issembayeva Anargul Shaushenova Ardak Nurpeisova Aidar Ispussinov Buldyryk Suleimenova Anargul Bekenova Aliya Satybaldieva Aigul Zholmukhanova Galiya Mauina

In this paper, we propose a spatiotemporal approach for binary classification of violent and non-violent behavior in real-world settings. The experimental pipeline includes video preprocessing, stratified data splitting, generation of temporally structured clips, and comparative evaluation of baseline models, including a convolutional neural network. We also developed a Residual Adaptive Motion Temporal Binary Heat Network model that combines frame color characteristics, residual motion descriptions, temporal feature fusion, an early risk assessment mechanism, and interpretable localization maps. Experiments were conducted on a balanced dataset of 2000 video clips. The proposed model demonstrated the best early warning performance: a supervision rate of 0.6, an F1 score of 0.9527, and a balanced accuracy of 0.9533. With full supervision, the F1 score was 0.9342, and the area under the receiver operating characteristic curve (AUC) was 0.9871. The practical significance of the work is that the proposed approach can be used as a decision support tool for the preliminary identification of potentially dangerous video fragments with subsequent manual verification, without the assumption of autonomous use in high-risk scenarios.

Computers doi: 10.3390/computers15050266

Authors: Abdallah Elayan Michel Kadoch

The exponential expansion of computer networks, alongside the rapid development of the Internet of Things (IoT), has significantly increased the volume and complexity of transmitted data, emphasizing the need for robust network security measures to secure sensitive data and prevent unauthorized access or breaches. Intrusion Detection Systems (IDSs) have emerged as a vital tool for protecting networks and IoT environments from threats. Various IDSs have been proposed in the literature; however, the lack of optimal feature learning, computational efficiency, and reliance on obsolete datasets poses significant challenges, limiting their effectiveness against evolving cyber threats. Moreover, traditional IDSs struggle to efficiently manage the high-dimensional and imbalanced nature of IoT network traffic data. To address these challenges, this research proposes a hybrid deep learning (DL)-based IDS integrating Binary Particle Swarm Optimization (BPSO), MultiHead Attention mechanisms (MHA), and a deep Gated Recurrent Unit (GRU) architecture, improving detection effectiveness while reducing computational overhead. Our proposed approach also utilizes a Target Sampling strategy to balance class distributions, enhancing the model’s ability to accurately identify minority attacks. The BPSO algorithm is employed to identify the most influential features from the high-dimensional network traffic datasets, enhancing model interpretability and supporting more efficient learning. This optimized feature subset is then fed into a GRU-based DL architecture augmented with MHA, which performs sequence processing and attention-based learning for intrusion detection. The performance of the proposed model is evaluated utilizing the BoT-IoT and the CIC-IDS2017 benchmark datasets, ensuring a comprehensive assessment of anomaly detection capabilities. Extensive experimental results demonstrate the superior performance of the proposed model, achieving a recall of 98.42% and 99.76%, with F1-score of 98.94% and 99.76% for binary classification and a recall of 99.79% and 98.69%, with F1-score of 99.89% and 98.04% for multiclass classification on the BoT-IoT and CIC-IDS2017 datasets, respectively, highlighting the effectiveness of our model in enhancing threat detection for computer networks and IoT environments in comparison to recent state-of-the-art IDSs.

Computers doi: 10.3390/computers15050265

Authors: Ana Chacón-Luna Miguel Tupac-Yupanqui Nicolás Márquez Cristian Vidal-Silva

Model-Based Diagnosis (MBD) is widely used to identify minimal conflicts and repair actions in constraint-based systems. Recent advances in parallel reasoning have significantly reduced runtime in large-scale models through speculative and multicore execution strategies. However, existing approaches primarily focus on computational efficiency and implicitly assume that minimal diagnoses are inherently suitable explanations for human decision makers. In complex configuration environments, minimality does not necessarily imply interpretability, as diagnoses may involve structurally dispersed or semantically heterogeneous constraints. To address this limitation, this paper introduces a multi-objective explainability-aware framework for parallel MDB. Diagnosis selection is formulated as a Pareto optimization problem balancing total computational cost and a formally defined interpretability penalty. Interpretability is quantified using graph-based structural dispersion, semantic entropy, hierarchical complexity, and ambiguity metrics. The proposed E-ParetoDiag algorithm computes non-dominated diagnoses and identifies balanced knee-point solutions without modifying correctness guarantees of underlying diagnosis algorithms. Experimental evaluation on large-scale benchmark datasets demonstrates a measurable trade-off between runtime and interpretability, particularly in dense constraint systems. Comparative analysis against classical selection strategies shows that the proposed approach reduces structural dispersion by up to 18% while increasing computational cost by only 7%. Statistical validation confirms that these improvements are significant (p < 0.01) in medium- and high-density scenarios. The results indicate that aggressive parallelism may improve computational efficiency while increasing explanation complexity, highlighting the need for multi-objective selection strategies. Overall, the proposed framework extends scalable symbolic reasoning toward a human-centered diagnosis paradigm and establishes a principled foundation for explainability-aware optimization in constraint-based systems.

Computers doi: 10.3390/computers15050264

Authors: Swapnaneel Dhar Riyanka Manna Khaldi Amine Aditya Kumar Sahu

This work introduces a non-blind watermarking framework for color images to address tamper detection, particularly under geometric transformations. The proposed scheme fuses two watermarks, a personal signature and a biometric fingerprint, into a unified composite watermark embedded into the chrominance component of the cover image using a multi-level transform domain approach, discrete wavelet transforms (DWTs), discrete cosine transforms (DCTs), and singular value decomposition (SVD). By leveraging the rotation-invariant properties of scale-invariant feature transform (SIFT) and oriented FAST and rotated BRIEF (ORB) descriptors, the framework ensures robust tamper detection without requiring alignment, thus mitigating the limitations of conventional detection techniques vulnerable to transformation-induced tamper obfuscation (TITO). Extensive experimentation demonstrates that the method maintains high perceptual fidelity, achieving PSNR values ranging from 50 to 55 dB for embedding strength factor μ (0.01–0.04) and SSIM indices near 1 across multiple benchmark images. Furthermore, the scheme exhibits notable resilience to a range of image processing attacks and geometric distortion. Comparative evaluation reveals its superiority over existing grayscale, color, SIFT-based and DWT-DCT-SVD-based watermarking techniques, affirming its applicability in scenarios demanding secure, imperceptible, and transformation-invariant image watermarking.

Computers doi: 10.3390/computers15050263

Authors: Adisak Intana Kuljaree Tantayakul Pongsakorn Kaewnaka

Software testing is essential for software engineering practices, as it ensures that the final software product is reliable and satisfies all requirements before delivery. However, manually designing black-box testing test cases is time-consuming, inconsistent, and difficult to maintain in accordance with changing specifications. Therefore, this paper presents BlackBoxTestGen, an automatic framework that unifies three specification-driven black-box testing techniques, including rule-based Equivalence Class Partitioning (ECP), syntax, and state transition testing. The framework utilises a redesigned XML structure for test case generation to be shared among a data dictionary, decision tree, and state machine, used by each testing technique. The degree of testing coverage is accumulatively calculated during the test case generation process. The beneficial value of our proposed framework was demonstrated with the development of a web-based prototype tool. We rigorously evaluated its performance in terms of accuracy, computational efficiency, and scalability through a multidimensional approach. This included assessment by professional experts, algorithmic stress testing via parameter scaling, and application to close-to-realistic case studies. The results indicate that BlackBoxTestGen provides a robust integration of testing techniques. By automating the generation of compact and reproducible test cases, the framework substantially reduces manual effort and minimises drift between techniques.

Computers doi: 10.3390/computers15050262

Authors: Constantinos Djouvas Christiana Andreou Maria C. Voutsa Nicolas Tsapatsoulis

Large Language Models (LLMs) are increasingly deployed as automated annotators in semantic multimedia systems, yet their reliability varies significantly across architectures. This study extends prior cross-model evaluations by benchmarking ChatGPT-5, Qwen-3, and Gemini-3-flash against human expert annotations using the HRAST hotel review dataset. We adopt a bias-by-design framework to analyze systematic divergences in sentiment, topic, and aspect labeling across real and synthetic data, while investigating the moderating effects of annotation mode. Findings reveal model-contingent polarity bias: ChatGPT-5 exhibits a pronounced neutrality bias, while Qwen-3 and Gemini-3-flash align more closely with human polarization. Agreement is substantial for concrete topics but diverges on abstract evaluative dimensions. Synthetic data consistently inflates reliability metrics while masking ambiguity. These findings highlight that annotation bias is structurally embedded in model design choices and operational conditions. Cross-architectural triangulation and mode-aware deployment strategies are recommended for robust semantic multimedia system development.

Computers doi: 10.3390/computers15050261

Authors: Qiuyang Sun Shaonan Liu Hong Li Yingchao Feng Liuqing Sun Kun Lu Kangtai Liu

Extreme low-light image enhancement (ELLIE) targets the restoration of visual quality under ultra-dim environments (<0.1 lux). Conventional image signal processing (ISP) pipelines often fail in such scenarios due to the limitations of heuristic, hand-crafted algorithms. While deep learning has advanced the field via end-to-end mapping, existing models suffer from constrained generalization and suboptimal perceptual fidelity, primarily stemming from the scarcity of large-scale, high-diversity datasets. To bridge this gap, we present the Day and Night (DaN) dataset, a semi-synthetic benchmark synthesized through a rigorous physics-based noise model. This approach effectively captures authentic noise characteristics while enabling the scalable generation of paired samples across multifaceted illumination conditions and scenes. Furthermore, we propose No Longer Vigil (NLV), a fully differentiable AI-ISP framework. By replacing traditional rigid blocks with adaptive non-linear networks, NLV facilitates scene-dependent transformations without requiring manual priors. Comprehensive evaluations demonstrate that our method significantly outshines state-of-the-art approaches, yielding a 4.15 dB gain in PSNR and a 0.026 improvement in SSIM.

Computers doi: 10.3390/computers15040260

Authors: Dina Shaltykova Akhat Bakirov Anastasiya Grishina Mariya Kostsova Yelizaveta Vitulyova Ibragim Suleimenov

This study examines the applicability of coding-theoretic tools to the digital monitoring of information space. The proposed approach treats response patterns to socially significant stimuli as binary sequences and interprets their analysis as a classification problem analogous to error correction in coding theory. To verify the feasibility of this framework, a model psychological test consisting of seven binary questions was analyzed using a procedure derived from the Hamming code (7,4). The method makes it possible to map the full space of observed answer combinations onto a smaller set of reference codewords and thereby identify stable response configurations. The obtained results show that the distributions produced after coding-based transformation are markedly non-uniform and contain recurrent maxima, indicating the presence of structured patterns in collective responses. It is also shown that permutations of question order substantially affect the resulting distributions and correlation indicators, which highlights both the sensitivity and the analytical potential of the proposed encoding scheme. The main contribution of the study is methodological: it demonstrates that error-correcting coding can be operationalized as a formal tool for detecting latent regularities in simplified monitoring data. At the same time, the present results should be regarded as proof of concept, since further work is required to validate the approach on larger datasets, compare it with baseline classification methods, and extend it to longer and multivalued response sequences.

Computers doi: 10.3390/computers15040259

Authors: Aiman Akynbekova Ayagoz Mukhanova Raikhan Muratkhan Lunara Diyarova Saya Baigubenova Gulden Murzabekova Gulaim Orazymbetova Aliya Satybaldieva Zhanat Abdikadyr

This study proposes a reproducible hybrid computational model for the explainable classification of territorial vulnerability using heterogeneous tabular data. The approach integrates fuzzy logic and extreme gradient boosting in a two-stage architecture that balances interpretability and predictive performance. First, a fuzzy transformation is applied to construct interpretable risk and resilience indicators based on multi-source administrative indicators. The analytical dataset was formed by integrating 11 heterogeneous administrative sources into a single matrix of 166 territorial units and 76 features. The model was evaluated on a stratified 75/25 split of the training and test sets using the F1 score, ROC-AUC, precision, recall, and integrated quality criterion. Experimental results show that the proposed Fuzzy-XGBoost framework achieved an F1 score of 0.7333 on the test dataset, an ROC-AUC of 0.8291, and an Integrated Score of 0.768, outperforming the strongest baseline and improving recall in highly vulnerable areas. Furthermore, probabilistic threshold optimization identified an operating point at τ = 0.35, reducing the number of missed high-risk cases while maintaining acceptable specificity. The results demonstrate that fuzzy feature expansion combined with gradient boosting provides an efficient and interpretable solution for tabular risk classification and decision support problems under heterogeneity and uncertainty.

Computers doi: 10.3390/computers15040258

Authors: Raluca Portase Eugen-Richard Ardelean

In recent years, deep learning has been applied more and more to medical image analysis. One such application of deep learning is the automated polyp detection in colonoscopy with the target of reducing miss rates. This study presents a comprehensive evaluation of nine state-of-the-art object detection models for colonic polyp detection: YOLOv8, YOLOv9, YOLOv10, YOLO11, YOLO12, YOLO26, RT-DETR, YOLO-World, and YOLOE. The models were evaluated on three publicly available datasets: CVC-ClinicDB, CVC-ColonDB, and ETIS-LaribPolypDB. All models were trained under standardized conditions using identical hyperparameters and data augmentation strategies to guarantee fair comparison. Performance was evaluated using multiple metrics: mAP@50, mAP@50–95, F1 score, precision, recall, inference time, and computational cost. YOLO11 demonstrated the best overall performance, achieving mAP@50 scores of 0.995, 0.944, and 0.978 on the three datasets respectively, while maintaining the fastest inference time of approximately 150 ms per image and the third-lowest computational cost at 21.3 GFLOPs. Cross-dataset generalization experiments revealed a significant loss of performance, with mAP@50 dropping by 20–40% when models were tested on an unseen dataset, highlighting the challenge of true generalization with limited datasets. Statistical analysis by polyp size showed that while all models achieved F1 scores exceeding 0.95 for large polyps, performance decreased to 0.60–0.85 for small polyps, indicating a limitation in detecting small lesions. The analysis of failure modes showed that missed detections, false positives and boundary errors constitute 60–75% of all failures, suggesting that domain adaptation of object detection models may be required.

Computers doi: 10.3390/computers15040257

Authors: Ahmad Salehi Shahraki Hooman Alavizadeh

Online cybersecurity education increasingly serves diverse cohorts, including students with non-technical backgrounds and those balancing their studies with work or family responsibilities. Yet, research on sustaining educational quality while scaling fully online enrolments remains limited, particularly in foundational technical subjects where learning requires both conceptual understanding and professional judgement. This study aims to examine how teaching presence can be operationalised in fully online foundational cybersecurity subjects through inspectable artefacts and routines that remain workable for large cohorts and distributed teaching teams. This paper reports a Scholarship of Teaching and Learning (SoTL) design and transfer case grounded in the Community of Inquiry (CoI) framework. This study examines the redesign of CSE1ICB (Introduction to Cybersecurity) and the transfer of the same design logic to CSE1CPR (Cybersecurity in Practice). The findings identify a coherent four-component design model comprising (1) real-world incident integration, (2) scenario-based learning and interactive checks, (3) structured, layered support, and (4) a predictable communication rhythm across the learning management system (LMS) and email. Across these two subjects, these elements are presented as an integrated system intended to make learning objectives salient, increase opportunities for guided practice in professional reasoning, reduce avoidable friction in practical work, and create consistent instructor visibility through routine communication and support structures. This paper synthesises the approach into nine transferable design principles, mapped to CoI teaching presence dimensions and illustrated through concrete design choices, including incident-framing templates, scenario prompt patterns, layered support resources, formative feedback patterns, and communication routines. Overall, this study shows that teaching presence can be operationalised as a coordinated design system rather than as a set of isolated tactics. This paper contributes a reusable and theory-informed model for educators coordinating foundational cybersecurity subjects delivered online at scale.

Computers doi: 10.3390/computers15040256

Authors: Mohamed Hanafy Abdel-Kader Mahmoud Sherine Nagy Saleh Amin Shoukry Yousry Elgamal

Emotions are fundamental to the human experience and are increasingly analyzed in applications such as marketing, healthcare, and human–computer interaction. Many recent approaches to human emotion recognition rely on deep learning, which typically demands large labeled datasets and substantial computational resources and often suffers from limited interpretability. Applying classical machine-learning methods to sensor time series is more lightweight but may struggle to reach high accuracy, especially when the temporal structure is not explicitly modelled. This paper introduces three subinterval voting-based ensemble models designed for user-specific emotion classification from multimodal time-series data acquired by smartwatch inertial sensors and heart-rate measurements. Each model partitions a time window into subwindows and performs window-level voting, thereby exploiting the temporal consistency of emotional responses while remaining compatible with standard classifiers such as logistic regression and Random Forests (with or without hyperparameter tuning). The models are evaluated on a public smartwatch emotion benchmark dataset under both binary (happy vs. sad) and three-class (happy, sad, neutral) settings. The relative accuracy improvement over the corresponding baseline reported in prior work ranges from 4.68% to 26.05%, with a mean gain of 12.34%. For the three-class tasks, improvements range from 11.17% to 37.10%, with a mean gain of 21.63%. Within the evaluated experimental setting, these results show that the proposed subinterval ensembles consistently enhance performance while remaining model-agnostic and compatible with standard user-specific classification pipelines in sensor-based emotion recognition.

Computers doi: 10.3390/computers15040255

Authors: Saadi Turied Kurdi Luttfi A. Al-Haddad Zeashan Hameed Khan

This study proposes a scalable, AI-driven, and Geographic Information System (GIS)-integrated framework for intelligent route-level classification in large-scale airport transportation networks to support airport operations, logistics planning, and network-level decision-making. The framework addresses the need for practical artificial intelligence applications that combine spatial network analysis with supervised machine learning to improve route assessment and resource allocation in complex air transport systems. A structured dataset was developed using operational and traffic-related attributes, including route distance, aircraft capacity, weekly frequency, annual passenger volume, demand variability, and route performance indicators, with additional normalized features to improve data representation. A Gradient Boosting ensemble classifier was trained to categorize routes into high-, medium-, and low-priority classes. The model achieved strong predictive performance, with a testing area under the ROC curve of 0.961, accuracy of 0.922, F1-score of 0.915, precision of 0.918, and a recall of 0.922. Feature importance analysis identified demand variability and route-density indicators as the main drivers of classification, enhancing interpretability and practical trust. The proposed framework demonstrates the real-world potential of AI for scalable, explainable, and efficient decision support in airport logistics and transportation network management.