Agriculture is a vital source of revenue and employment in developing nations like Bangladesh, yet traditional farming techniques often lead to sub-optimal productivity. While modern precision agriculture systems offer data-driven solutions, they frequently rely on computationally intensive models that are incompatible with the limited hardware of low-end smartphones and rural IoT devices. This research proposes a lightweight, modular hybrid framework for optimal crop selection, yield estimation, and input optimization designed to bridge this digital divide. The architecture employs a Gaussian Naive Bayes (GNB) classifier for crop recommendation, achieving a remarkable accuracy of 99.55% while maintaining an ultra-lightweight memory footprint of approximately 3.4 KB-which is 196.1 times smaller than traditional Random Forest models. The second stage utilizes a Decision Tree Regressor (DTR) to provide numerical yield estimates with an $R^{2}$ score of 0.9517. To go beyond mere prediction, the framework integrates a Genetic Algorithm (GA) layer that identifies optimal nutrient and environmental configurations, leading to a predicted yield improvement of 12% to 18%. The entire system is designed for offline execution on resource-constrained devices, ensuring sub-millisecond inference times. This scalable, on-farm solution empowers marginalized farmers in remote regions, such as the Chattogram Hill Tracts, to make high-precision, prescriptive agricultural decisions on affordable hardware.

Corn leaf diseases significantly affect crop yield and quality, creating serious challenges for sustainable agriculture. Although deep learning approaches, particularly Convolutional Neural Networks (CNNs), have achieved success in automated plant disease classification, their effectiveness is limited by the requirement of large-scale labeled datasets, which are costly and time-consuming to obtain. To overcome this limitation, we propose a framework that leverages both Self-Supervised Learning (SSL) and Semi-Supervised Learning (Semi-SL) to learn robust visual representations from largely unlabeled corn leaf images. We employ state-of-the-art SSL methods, including SimCLR and MoCo, with lightweight classifiers such as logistic regression and support vector machine (SVM) for downstream disease classification under label-scarce conditions. Additionally, we adopt FixMatch as a semi-SL method, exploring the impact of pseudolabeling confidence thresholds on classification performance. Our experiments demonstrate that SimCLR with logistic regression achieves the highest accuracy of 92.08% with 10% downstream data, while FixMatch with 10% labeled data and a threshold of 0.8 achieves the best overall results, with 93.25% F1 macro and 94.82% accuracy. These results confirm that SSL and semi-SL approaches can achieve competitive performance while substantially reducing dependence on labeled data, offering a scalable and cost-effective solution for early detection and management of corn leaf diseases.

Unauthorized drone activity in restricted areas such as airports, military facilities, and critical infrastructure sites creates increasing security concerns. Visual drone detection using deep learning offers a practical solution, but its performance usually depends on large amounts of annotated training data. This paper presents a label-efficient framework for single-class DJI Tello drone detection by combining YOLO-based object detection with self-supervised and semi-supervised learning strategies. An open-source dataset containing 860 DJI Tello quadcopter images from indoor and outdoor environments is used for evaluation. YOLOv10, YOLOv11, and YOLOv12 are first benchmarked under full supervision, where YOLOv11 achieves the best supervised performance with mAP@0.50 of 0.9930 and recall of 0.9922. The selected YOLOv11 backbone is then evaluated with three label-efficient strategies, namely Masked Autoencoder (MAE), DINOv2, and Pseudo-STAC, using only 20% labeled data. Among these methods, MAE achieves the highest mAP@0.50 of 0.9922, recall of 0.9690, and F1 score of 0.9793, closely approaching the fully supervised baseline while reducing annotation requirements. Runtime analysis further shows that the MAE-enhanced YOLOv11n detector achieves 12.98 ms per frame and 77.06 FPS on a Tesla T4 GPU. These results indicate that self-supervised pre-training can improve label-efficient drone detection while maintaining real-time inference performance for restricted-area surveillance applications.

Fake news is false or misleading information shared as real news, often meant to trick or influence public opinion. Its fast spread through social media causes confusion, divides communities, and reduces trust in reliable news sources. While much research has been done on fake news detection in English, there is a lack of tools designed specifically for Bangla using a combined linguistic and psychological approach, despite the rise of fake news in Bangla-speaking regions. Our research aims to bridge this gap by developing tailored tools that effectively address the nuances of fake news in the Bangla language. By implementing a robust detection model, we contribute to a deeper understanding of misinformation in this context. In this work, we present a psycholinguistic method that combines linguistic and psychological features of the Bangla language to detect fake news. We created a dataset of 5,428 Bangla news articles, categorized as real or fake, capturing diverse topics and styles. This research highlights the importance of developing language-specific tools to combat misinformation. It also offers a valuable resource for future studies, helping to promote a more informed and responsible media environment in Bangla-speaking communities. Our hybrid model, which integrates Convolutional Neural Networks (CNN) and Multilayer Perceptron (MLP), effectively captures these features, achieving an accuracy of 96.32%. These results demonstrate the strong potential of our approach in improving the detection of fake news in Bangla.

Cardiovascular diseases remain a leading global health concern, necessitating accurate, non-invasive blood pressure monitoring solutions. This study presents a novel photoplethysmography (PPG)-based system that integrates three key innovations: (1) a real-time signal quality assessment (SQA) model employing a one-class SVM classifier to filter unreliable PPG segments, (2) a computationally efficient two-stage neural network (1D U-Net followed by 1D MultiResUNet) for arterial blood pressure (ABP) waveform estimation, and (3) a wearable hardware platform combining a MAX30102 optical sensor and ESP32 microcontroller for portable deployment. The SQA model uses five optimized features to achieve robust motion artifact rejection. The deep learning pipeline reconstructs ABP waveforms with mean absolute errors of 4.7 mmHg (systolic) and 4.3 mmHg (diastolic), complying with the Association for the Advancement of Medical Instrumentation (AAMI) standards. Validated on 40 subjects against reference sphygmomanometer measurements, the system demonstrates 92.5% (systolic) and 95% (diastolic) classification accuracy, with subsecond latency and low-power operation. By addressing critical challenges in motion robustness, computational efficiency, and clinical validation, this work advances the practicality of cuffless BP monitoring for telehealth and resource-limited settings.

Smishing (SMS phishing) is a form of social engineering attack in which adversaries exploit mobile messaging platforms to deceive users into disclosing sensitive information, installing malicious content, or initiating fraudulent transactions. Unlike traditional cyberattacks that target software vulnerabilities, smishing primarily exploits human trust, making it difficult to detect using conventional rule-based or signature-driven security mechanisms.With the rapid growth of mobile communication and digital financial services, smishing has emerged as a significant cybersecurity threat affecting individuals, enterprises, and communication infrastructures. To address this challenge, this paper proposes a cybersecurity-oriented hybrid fusion framework for smishing detection that integrates advanced feature engineering with deep learning and transformer-based language models. The proposed approach combines semantic embeddings from Sentence-BERT, statistical TF–IDF representations, and SMS-specific security features capturing URL behavior, phone number patterns, sentiment manipulation, linguistic anomalies, and named entity indicators commonly associated with phishing attacks. These heterogeneous features are fused through a multilayer feed-forward neural network, enabling the detection of both known and previously unseen smishing strategies. The framework is evaluated on six publicly available SMS datasets containing English and Bangla messages, ranging from 2,772 to 16,572 samples after preprocessing and class balancing. Experiments follow a stratified train–validation–test split (70%–15%–15%) with 5-fold cross-validation and Optuna-based hyperparameter optimization. The proposed model achieves a peak validation accuracy of 99.67%, an average cross-validation accuracy of 99.38%, and a test accuracy of up to 96.54%, with ROC–AUC values exceeding 98%. The results demonstrate that the proposed approach outperforms traditional machine learning, deep learning, and standalone transformer-based models. To enhance model transparency, Local Interpretable Model-agnostic Explanations (LIME) are incorporated to provide interpretable insights into feature-level decision behavior, demonstrating the effectiveness of the proposed framework for real-world smishing detection in multilingual mobile communication environments.

The rapid transition to online learning has intensified the need to understand the multifaceted drivers of student satisfaction in virtual education environments. Student satisfaction is a cornerstone of effective digital education, influencing academic performance, learner engagement, and institutional decision-making. As a critical predictor of retention and perceived instructional quality, it directly impacts the sustainability of online learning systems. However, existing studies often lack integrated frameworks that combine subjective expert judgment, global and local model interpretability, and subgroup-specific analysis. To address this gap, this study aims to develop a hybrid decision-analytic and explainable machine learning framework to systematically assess and interpret student satisfaction in virtual classrooms. The proposed framework adopts a layered analytical design in which statistical feature relevance (ANOVA), model-based interpretability (SHAP and LIME), and expert-driven weighting under uncertainty (FAHP) are treated as complementary but non-overlapping components, with TOPSIS used exclusively for decision-level satisfaction ranking. A structured survey has been conducted among 1,469 university students, capturing responses across 21 features encompassing demographic attributes, digital readiness, behavioral patterns, and perceived instructional quality. Feature selection has been performed using the ANOVA F-test, while SHapley Additive Explanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) have been applied to derive global and local feature importance. To incorporate subjective expert judgment, feature weights have been calculated using the Fuzzy Analytic Hierarchy Process (FAHP), and composite satisfaction scores have been computed using the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS). These scores have been used to segment students via Agglomerative Hierarchical Clustering into low (n = 852), medium (n = 484), and high (n = 133) satisfaction cohorts. Cluster-specific SHAP and LIME analyses have uncovered distinct patterns of feature influence across subgroups. The identified satisfaction drivers align closely with constructs from established models of technology acceptance and learning motivation, including the Technology Acceptance Model (TAM), the Unified Theory of Acceptance and Use of Technology (UTAUT), and Self-Determination Theory (SDT). High satisfaction has been linked to emotional positivity, perceived equivalence to offline instruction, and academic self-efficacy, whereas low satisfaction has been driven by affordability barriers, psychological stress, and misalignment with preferred learning modalities. The proposed framework offers a comprehensive, interpretable, and rank-aware approach to modeling student satisfaction, equipping educational stakeholders with actionable insights for designing more inclusive, equitable, and learner-centered online education environments.

Accidental falls have emerged as a major public health concern, especially among individuals aged 65 and older, due to their high incidence and severe consequences. Without timely intervention, such falls can result in fractures, traumatic brain injuries, and long-term complications. As a result, considerable research has focused on developing automated fall detection systems that integrate intelligent algorithms with sensor-based data acquisition to enable rapid response and medical assistance. This study follows the PRISMA framework to conduct a systematic literature review. A comprehensive search was performed across major databases including PubMed, Google Scholar, Scopus, and IEEE Xplore using fall detection–related keywords. From an initial pool of 596 articles, duplicates were removed and strict inclusion/exclusion criteria were applied, resulting in 182 relevant articles for in-depth analysis. This review examines a wide range of Artificial Intelligence (AI) and Machine Learning (ML) approaches applied to fall detection using diverse sensor modalities, including wearable, vision-based, ambient, and multimodal systems. Additionally, it summarizes the publicly available datasets and sensor configurations used in existing studies. This broad perspective is essential to compare trade-offs across sensing modalities and support effective system development for real-world deployment. This review provides a comparative overview of AI/ML-based fall detection approaches, highlighting differences in accuracy, sensitivity, and dataset usage as reported in existing literature. In the end, many open research challenges in using AI and ML to detect falls are outlined, along with potential future perspectives.

Brain tumor detection from MRI images plays a critical role in early diagnosis and effective treatment planning. Deep learning methods have shown promising results in medical image analysis, including brain tumor detection. In this paper, we propose a deep learning pipeline for accurate brain tumor detection in MRI images using the VGG-16 model. The proposed pipeline consists of multiple stages, starting with preprocessing to enhance image quality and reduce noise. The preprocessed images are then fed into the VGG-16 model, which has been pretrained on a large dataset of natural images, and fine-tuned on a specialized dataset of brain MRI images. The model leverages its deep architecture to automatically learn intricate features representative of tumor regions. To evaluate the effectiveness of the pipeline, extensive experiments were conducted on a diverse dataset of brain MRI scans. The results demonstrate that our approach achieved an impressive accuracy of 99.99

Prostate cancer, the second most prevalent male malignancy, requires advanced diagnostic tools. We propose an explainable AI system combining BERT (for textual clinical notes) and Random Forest (for numerical lab data) through a novel multimodal fusion strategy, achieving superior classification performance on PLCO-NIH dataset (98

Predicting the energy consumption of buildings plays a critical role in supporting utility providers, users, and facility managers in minimizing energy waste and optimizing operational efficiency. However, this prediction becomes difficult because of the limited availability of supervised labeled data to train Artificial Intelligence (AI) models. This data availability becomes either expensive or difficult due to privacy protection. To overcome the scarcity of balanced labeled data, semi-supervised learning utilizes extensive unlabeled data. Motivated by this, we propose semi-supervised learning to train AI model. For the AI model, we employ the Belief Rule-Based Expert System (BRBES) because of its domain knowledge-based prediction and uncertainty handling mechanism. For improved accuracy of the BRBES, we utilize initial labeled data to optimize BRBES’ parameters and structure through evolutionary learning until its accuracy reaches the confidence threshold. As semi-supervised learning, we employ a self-training model to assign pseudo-labels, predicted by the BRBES, to unlabeled data generated through weak and strong augmentation. We reoptimize the BRBES with labeled and pseudo-labeled data, resulting in a semi-supervised BRBES. Finally, this semi-supervised BRBES explains its prediction to the end-user in nontechnical human language, resulting in a trust relationship. To validate our proposed semi-supervised explainable BRBES framework, a case study based on Skellefteå, Sweden, is used to predict and explain energy consumption of buildings. Experimental results show 20 ± 0.71% higher accuracy of the semi-supervised BRBES than state-of-the-art semi-supervised machine learning models. Moreover, the semi-supervised BRBES framework turns out to be 29 ± 0.67% more explainable than these semi-supervised machine learning models.

The counterfeit medication infiltration within global supply chains poses a major public health threat. To address this, a collaborative effort among governments, regulators, and pharmaceutical companies is essential to secure the global/local supply chain. This paper proposes a novel approach that leverages blockchain technology, polymorphic encryption, and cloud storage to tackle security risks and privacy concerns in medication supply chains. The framework integrates a drug supply chain decentralized application (also called SCMapp) within the Ethereum blockchain, enabling functionalities like secure supplier onboarding, encrypted data management, cloud storage integration, and efficient data retrieval. This approach aims to revolutionize drug supply chain management by enhancing security, transparency, and overall efficiency, ensuring adherence to global health regulations. A safe and effective method for managing drug supply chains is provided by the suggested Drug Supply Chain Management System. The proposed model outperformed existing solutions in terms of security, efficiency, and traceability. The combination of encryption, blockchain, and cloud storage provided a comprehensive approach to address the challenges of drug supply chain management. The comparison analysis highlighted the unique advantages of the proposed model over other methods.

Edge computing is emerging as a transformative approach in Industrial Internet of Things (IIoT) applications, particularly within remote industrial environments where cloud connectivity is unreliable. This study investigates a fully edge-based data analytics solution using an open source platform to enhance real-time monitoring, alarm systems, and sensor health analysis without reliance on cloud infrastructure. The experimental setup utilizes Flasheye's LiDAR perception system (which monitors areas in underground mining) integrated with MQTT to process and visualize data on ThingsBoard, focusing on critical parameters such as throughput, latency, and failure rate. Evaluation results demonstrate that this edge-based approach supports low-latency performance and reliable alarm management, underscoring the practicality of localized analytics for remote industrial operations. By implementing and evaluating an open source edge computing solution in remote industrial environments, the study contributes a fully edge-based data analytics solution for IIoT deployments, enhancing operational efficiency and safety in connectivity-limited environments.

Malicious URLs represent a significant cybersecurity threat, facilitating malware distribution and data theft. This paper explores a ML-based framework for malicious URL detection, providing a comparative analysis of DL and traditional ML approaches. Eight ML models─LR, SVM, DT, KNN, GNB, RF, XGBoost, and LightGBM─are benchmarked against three DL models: LSTM, BiLSTM, and GRU. The results reveal that traditional ML models, particularly RF, XGBoost, and LightGBM, achieve superior performance with accuracy scores of up to 92%, outperforming DL models, which achieve accuracy rates of 90%, 91%, and 88%, respectively. To further enhance detection performance, a stacking model combining these techniques is proposed, achieving a remarkable accuracy of 99.99%. This research underscores the potential of stacked models to significantly improve malicious URL detection, offering advanced solutions to strengthen cybersecurity frameworks for both individuals and organizations.

The rapid expansion of electric vehicles (EVs) has added complexity to the resilience and security challenges to the EV charging systems, especially owing to the exposure to the cyber–physical threats and the reliance on centrally coordinated systems. Although the previous literature has discussed the use of blockchain in the context of smart grids and mobility services; its implementation to improve the resilience of EV charging, particularly when integrated with cybersecurity systems, is still insufficiently synthesized. Despite these issues, critical gaps persist in terms of scalability, interoperability, and cybersecurity enforcement. This study presents an exploratory literature review that examines the intersection of blockchain and cybersecurity enabled applications and introduces a comparative framework evaluating the conventional security controls with blockchain based cybersecurity solutions to improve the resilience of EV charging infrastructure. The authors analyzed 70 studies published between 2018 and 2025 to determine the security weaknesses and map them to decentralized solutions. Reported threats, security mechanisms, architectural decisions, and levels of validation were grouped and reviewed critically in the patterns of limitations with respect to scalability, interoperability, and deployment maturity. Through the synthesis of fragmented results in cross disciplinary research, the paper finds the main gaps in research and comparative research results that could be used as a comprehensive reference in future studies and system design in resilient EV charging infrastructures.

Speech Emotion Recognition (SER) is a complex endeavor in human-computer interaction (HCI) that necessitates the use of artificial intelligence and deep learning to accurately classify emotional states, which are determined by analyzing speech audio signals. Bangla is classified as a low-resource language for SER due to the scarcity of labeled datasets, despite its status as the seventh most frequently spoken language at the global level. By utilizing the SUBESCO and BanglaSER corpora, which are both audio-only Bangla emotive speech datasets, this study aims to enhance emotion recognition in Bengali speech. Noise was eliminated through the application of Envelope Masking during preprocessing, and Mel-Frequency Cepstral Coefficients (MFCCs) were extracted to capture critical spectral features. Machine learning models such as K-Nearest Neighbor (KNN), Random Forest, and Multi-Layer Perceptron (MLP) are implemented by the system, in addition to ensemble techniques like Voting and Stacking Classifiers, to optimize its performance. Further, in order to process temporal and sequential speech patterns efficiently, deep learning architectures such as Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), and Long Short-Term Memory (LSTM) were implemented. The models that were proposed exhibited a high level of perceptual efficiency, obtaining an accuracy of 95.92% on SUBESCO and 90.61% on BanglaSER. These findings substantiate the efficacy of the preprocessing techniques and applied learning models, thereby enhancing BanglaSER and broadening the scope of research opportunities for low-resource languages.

With the exponential growth of digital systems, traditional centralized access control mechanisms such as RoleBased Access Control (RBAC) and Attribute-Based Access Control (ABAC) face significant challenges, including single points of failure, scalability limitations, and vulnerabilities to insider threats. These issues compromise the security, transparency, and auditability of access policies. This article explores the integration of blockchain technology with access control frameworks to mitigate these vulnerabilities. By leveraging decentralized ledgers and smart contracts, blockchain enhances transparency, tamperresistance, and traceability in policy enforcement and auditing processes. The study compares RBAC and ABAC implementations within blockchain environments, highlighting improvements in granularity, interoperability, and resilience. Through architecture models, smart contract code, and performance discussion, we demonstrate the potential of blockchain in enhancing access control security, transparency, and trustworthiness.

In the realm of digital communication and cybersecurity, the identification and filtering of ham and spam messages pose significant challenges due to the overwhelming volume of unsolicited and unwanted emails. This paper presents an in-depth analysis of various ML and DL techniques for efficient and robust spam detection systems, critical for enhancing cybersecurity defenses. Seven ML algorithms-RF, LR, SVM, XGBoost, GB, NB, and KNN-along with four deep learning models-CNN, LSTM, BiLSTM, and RNN-are evaluated for their effectiveness in spam classification. Additionally, we fine-tuned the BERT model, achieving a ground breaking accuracy of 99.37%, surpassing the 99.14% accuracy of the Bidirectional and Auto-Regressive Transformers (BART) model reported in recent research. Using a publicly available dataset of labeled ham and spam messages, the models were trained and tested, and their performance was assessed based on accuracy, precision, recall, and F1 score. The results demonstrate the superiority of the BERT model in spam detection, setting a new benchmark for cybersecurity. The success of BERT is attributed to its advanced capability to capture intricate patterns and contextual information, essential for distinguishing legitimate messages from spam, thus bolstering cybersecurity efforts.