This study addresses the challenge of securing smart-home Internet-of-Things (IoT) systems under severe resource constraints by proposing and evaluating a lightweight hybrid framework that couples on-device anomaly detection (TinyML) with elliptic-curve cryptography (ECC) and blockchain-based event logging. The approach first classifies incoming sensor readings locally using a TinyML anomaly detector (Isolation Forest); normal data are then encrypted with ECC and transmitted, while all security relevant actions are immutably recorded on a blockchain ledger to provide auditability and device trust. The framework was implemented on a smart home dataset of 49,000 records. The TinyML model achieved strong detection performance (0.98 Precision, 0.97 Recall, 0.975 F1-score, 0.996 Accuracy). Cryptographic and logging overheads were small average ECC key generation in 5.12 ms, encryption 0.85 ms, decryption 0.82 ms and blockchain logging. Overall, the results indicate that combining on device anomaly detection with ECC-secured communication and tamper-evident logging can deliver end-to-end protection, transparency, and scalability for smart-home IoT.

The errors of human annotation and the noise of the environment such as lighting changes, occlusions and cluttered backdrop limit the correct detection of the plant diseases in the field condition. The research hypothesis is to present a robust deep learning model that can withstand noise and be interpretable in controlled and noisy environments to achieve high plant disease classification. The hybrid EfficientNet-Vision Transformer (ViT) network proposed is based on an EfficientNet-B4 branch of CNN and a branch of Vision Transformer (ViT) network, which focuses on capturing fine-grained lesion features and global contexts information. A data augmentation pipeline based on CycleGAN is used to introduce field-style distortions (e.g., (lighting shifts, shadowing, debris and partial occlusions), to be more robust to environmental noise, and an Adaptive Symmetric Cross-Entropy (ASCE) loss identifies and down-weights uncertain samples with normalized prediction entropy. The training is done in two phases, Stage 1 pretraining with clean images of PlantVillage and Stage 2 with increasingly noisy samples. The framework is tested in two different noise conditions, and these include the controlled synthetic label noise with PlantVillage and the real environmental noise with PlantDoc. The proposed model has an accuracy of 94.5% on the clean PlantVillage test set. It achieves 85.0% accuracy on the PlantVillage dataset under the 20% synthetic label noise protocol, outperforming ResNet-50V2 (76.5%), DenseNet-121 (78.9%), and Co-Teaching (79.5%). Macro-precision, macro-recall and macro-F1 of the model on the external PlantDoc field dataset are 0.718, 0.681, 0.681, respectively with a top-1 accuracy of 72.0, which is a manifestation of cross-domain generalization. The lesion-centric Grad-CAM images indicate that the model places emphasis on symptomatic areas of leaves and represses reactions of background soil, shadows, and clutters. The suggested hybrid EfficientNet-ViT architecture offers, in general, a robust and explainable solution to precision agriculture and intelligent crop tracking systems that are resistant to noise.

This study presents a comprehensive analytical-applied framework for the revitalization of historic urban centers through the application of an urban injection strategy guided by the results of spatial structure analysis. The historic Rusafa district of Baghdad serves as a case study. Field surveys, digital maps, and Depth map X analysis were employed to measure indices of integration, clarity, intensity, and selection. This allowed for the identification of areas with high structural coherence, visually prominent axes, and centers of urban activity. The results reveal a dual urban structure within the historic core. This structure is characterized by highly integrated and spatially dominant main axes, such as Al Khulafa Street and Al-Khilani Square, contrasted with secondary streets and pathways suffering from weak coherence and functional performance, particularly sections of Al Rashid Street. Activity is clearly concentrated along the Al-Rashid-Al-Khulafa axis. Based on this diagnosis, the study proposes the Urban Injection strategy as a precise intervention approach. It draws inspiration from therapeutic and aesthetic intervention patterns in the medical field (intravenous, intramuscular, subcutaneous, intraosseous, cosmetic, preventative, and spiritual injections) and repurposes them within the field of architecture and urban planning, adapting them to the spatial functioning of the historic fabric. This approach allows for targeted interventions that revitalize urban areas while preserving heritage values. It provides planners and decision-makers with an evidence-based and replicable tool that supports the achievement of the Sustainable Development Goals, particularly Goal 11 (Sustainable Cities and Communities) and Goal 13 (Climate Action). The study is constrained by limited socio-economic data and restricted access to parts of the study area. Moreover, the analysis focuses mainly on spatial structure, with limited consideration of social and environmental variables. Future research should incorporate behavioral and environmental indicators to broaden the framework’s analytical scope.

Automatic anomaly detection in video surveillance is crucial for public and private safety. However, it is challenging because of unclear abnormal events, limited labeled data, and mismatches between different types of data. Traditional video anomaly detection methods mainly focus on spatiotemporal visual features. They often ignore semantic information and interactions between different data types. Additionally, many multimodal approaches use basic fusion methods that do not solve the alignment problems between these types of data. To address these issues, we propose a multimodal framework that includes a Hierarchical Multi-scale Temporal Network (H-MSTN). This network models short-, medium-, and long-term dependencies in visual and textual data. A lightweight cross-modal attention module makes sure the semantics align. Meanwhile, a Multimodal Attention-Based Fusion Transformer (MAFT) refines cross-modal representations in real time. We evaluate this framework using the UCF-Crime and XD-Violence benchmarks. The proposed method achieves 92.42% AUC on UCF-Crime and 88.63% AP on XD-Violence with significantly lower computational cost and faster inference than recent multimodal baselines such as ReFLIP-VAD. These results demonstrate a strong efficiency–accuracy trade-off for real-time deployment while maintaining competitive or improved performance over prior methods such as MVAD and TEVAD.

The event-related potential (ERP) analysis allows us to measure brain activity as it reflects sensory processing, attention, and a range of cognitive tasks. In this research article we introduce a comprehensive ERP-based approach for the identification of neurophysiological markers of alcoholism, using the open EEG dataset and the MNE-Python toolkit. All the EEG data recorded during a visual object recognition task were uniformly processed for both the alcoholic and control groups. After filtering and re-referencing, independent component analysis was applied, followed by segmenting the data into epochs and performing baseline correction. We focus on well-known ERP components, notably N2 and P300, occurring roughly 200-300 ms and 300-600 ms after the onset of the stimulus, respectively, which are associated with cognitive evaluation processes. We clearly see two distinct ERP profiles between the two groups. The alcoholic group shows reduced P300 and altered N2 compared with controls. This study presents a transparent and reproducible ERP analysis pipeline, developed solely with open-source tools and data, and highlights the potential of ERP markers as neurophysiological indicators of cognitive changes associated with alcohol use disorder.

Atherosclerosis is a progressive cardiovascular condition where arteries narrow due to plaque buildup, significantly increasing the risk of heart attacks and strokes. This study presents a non-invasive, Deep Learning-architecture based system for the timely diagnosis of atherosclerosis by means of real-time physiological and clinical data. This system integrates wearable sensors namely, Electrocardiogram (ECG), Photoplethysmography (PPG), Galvanic Skin Response (GSR), and Blood Pressure (BP) to continuously monitor heart-related parameters. Also, incorporates the clinical indicators namely blood glucose and cholesterol levels to enhance predictive accuracy. Data set comprises 226 records of the subjects having signs of atherosclerosis and 180 records of healthy subjects. The feature extraction involves total 8 original and 5 engineered features. Collected data undergoes preprocessing and is analyzed using various Deep Learning architectures including LSTM, BiLSTM, GRU, CNN-LSTM, and Transformer networks. These models are trained and evaluated using stratified K-fold cross-validation, ensuring consistent and generalized performance. The assessment metrics involves accuracy, precision, recall and F1 score. Among these, CNN-LSTM and Transformer models achieved superior accuracy and robustness in classifying individuals as healthy or at risk of atherosclerosis. The best model is chosen as CNN-LSTM with highest weighted score of 0.98 in comparison with other individual models. The final model is deployed in a user-friendly Streamlit interface, which helps users to input physiological data and receive real-time health predictions. The system provides a diagnostic output, confidence score, and highlights of any abnormal parameters. This solution addresses limitations of traditional diagnostics such as high cost, invasiveness, and lack of real-time feedback by offering a portable, affordable, and continuous monitoring tool. It empowers users, especially in remote or underserved areas, to take proactive measures for stroke prevention and cardiovascular health management.

The increasing demand for the Industrial Internet of Things (IIoT), with billions of connected things and the decentralization of data exchange, is gaining momentum, making conventional threat detection and analysis challenging in such distributed environments. In this paper, a security framework for edge nodes, called the Intrusion Detection, Prevention, and Response System (IDPRS), is proposed. It aims to detect MQTT (Message Queuing Telemetry Transport)-based threats using Machine Learning (ML) algorithms. However, ML models cannot be trained on resource-constrained devices; therefore, the approach trains the model on a high-performance platform, which will later serve as the detection engine on an edge node. The edge node can be hosted on low-cost single-board computers (SBCs), such as the Raspberry Pi. The detection model is further monitored and updated using an upgrade algorithm to make it adaptive to emerging threats. The evaluation results demonstrate high detection accuracy and reasonable resource and network overhead.

Deep learning has enhanced the analysis of medical images but privacy issues and institutional variations restrict their large scale application in clinics. FedVLM, a federated vision language model tailored to privacy-preserving multimodal medical image analysis, is one of the solutions to these problems. Contrary to the conventional federated design, which can only process single modal image data, FedVLM takes paired radiological images and clinical reports jointly, which demonstrates high zero-shot and few-shot diagnostic performance. The design consists of secure aggregation, differential privacy and proximal optimization that ensure protection of patient data and minimize variability across sites. Large scale experiments on the NIH ChestX-ray14, MIMIC-CXR, and BraTS datasets indicate that FedVLM is an accurate and interpretable model that achieves near-centralized performance on vision language models without violating privacy. Building on previous works such as FACMIC, BioViL, and FAA-CLIP, FedVLM introduces new methods, including privacy-aware optimization, proximal regularization for varied data, and multimodal contrastive alignment, creating a unified federated framework for clear and secure medical image analysis. Although FedVLM shows promising performance, this work is currently at a research stage and is not yet ready for clinical use. We need validation through future multi-institutional clinical studies.

This study experimentally investigates the influence of Al?O? nanoparticle addition on the combustion, performance, and emissions of emulsified Jatropha biodiesel in a compression-ignition engine. An emulsified fuel blend comprising 88% Jatropha methyl ester (JME), 10% water (v/v), and 2% surfactant (B20W10) was prepared using ultrasonication, into which Al?O? nanoparticles were dispersed at concentrations of 25 ppm and 50 ppm. Tests were conducted at varying loads under constant speed to evaluate performance, combustion, and emission characteristics. Among the tested fuels, B20W10Al50 yielded the best outcomes, achieving a 2.03% increase in brake thermal efficiency (BTE) and a 3.84% reduction in brake specific fuel consumption (BSFC) compared to diesel, with statistical analysis confirming the significance of these improvements. Combustion analysis showed a modest increase in peak in-cylinder pressure for B20W10Al50. Emission reductions were substantial relative to diesel: unburned hydrocarbons decreased by 40%, CO by 66.7%, NOx by 22.7%, and smoke opacity by 41.7%. These findings demonstrate that nanoparticle-assisted emulsification can address the common biodiesel trade-offs between efficiency and NOx formation. The study highlights B20W10Al50 as a promising formulation for sustainable transport applications, while also noting the need for further research on long-term nanoparticle stability, injector compatibility, and durability under real-world operating conditions.

Sentiment analysis utilizing artificial intelligence offers a transformative approach to managing health crises among diabetics in smart urban environments. This research proposes a practical AI-based solution that can be integrated into existing smart urban infrastructure to support real-time health crisis interventions for diabetic patients. Challenges in sentiment analysis for health crisis management in diabetics using AI include the need for high-quality, diverse data to accurately capture sentiment and the potential for privacy issues with sensitive health information in smart urban environments. The objective of this study is to leverage sentiment analysis utilizing artificial intelligence to enhance health crisis management for diabetics within smart urban environments. Adaptive Median Filtering Technique (AMFT) is used in pre-processing to reduce noise in sentiment analysis, as textual data from sources often contains noise such as irrelevant information, spam, and outliers. The combination of AMFT for noise reduction, RNNs for temporal sentiment analysis, and AI-driven optimization introduces a novel, technologically advanced approach to health crisis prediction systems. Recurrent Neural Network (RNN) models are highly effective for sentiment analysis, especially in the health crisis management of diabetics within smart urban environments, due to their ability to process sequential data and capture temporal dependencies. AI-driven optimization (AIDO) can automatically tune hyperparameters of sentiment analysis models in RNNs to improve performance, ensuring the models are both accurate and efficient. The AI-driven sentiment analysis system outperforms traditional monitoring methods, such as rule-based lexicons and keyword frequency-based approaches implemented in Python, achieving an accuracy of 0.92, a precision of 0.90, and a recall of 0.93.The proposed system reflects the focus on applied science and technological innovations by demonstrating a scalable, intelligent health monitoring framework that can be deployed in smart cities and urban health systems. Future advancements in sentiment analysis using artificial intelligence could enhance real-time monitoring and prediction of health crises in diabetics, integrating more diverse data sources and adaptive learning algorithms.