Remote patient monitoring, improved energy efficiency, besides the automation of mundane household duties are just a few samples of Internet of Things (IoT) has changed several industries. Attacks based on botnets, including distributed denial of service (DDoS), can compromise unprotected IoT devices. The use of deep learning (DL) to enhance privacy has been spurred by the realisation that traditional machine learning models employed to identify these threats violate data privacy. On the other hand, when it comes to cyberattack detection models based on DL, the majority of them fail to take computational complexity into account, making them unsuitable for placement on resource-constrained IoT edge devices. In order to identify botnet attacks on the IoT, this study presents a DL model that is computationally simple. The study uses dimensionality reduction and feature selection to keep computational complexity low and accuracy high. To begin, the botnet dataset’s best characteristics are chosen using an optimised Deep Belief Network (DBN) classical trained with repeated stratified k-fold cross-validation. In order to increase the classification accuracy, the study used Fire Hawk Optimisation (FHO) to fine-tune adjustable parameters. In comparison to earlier efforts, the model improved upon the curve. In smart grids, smart homes, and other contexts with deployed IoT edge devices that are resource constrained, the proposed methodology works well for botnet attack detection.

Red Palm Weevil (RPW) infestation poses a significant threat to coconut trees leading to severe damage and economic losses. However, existing RPW detection techniques have low sensitivity to early infestations as internal damage by larvae is not externally visible, hindering timely intervention. In this work, a novel deep learning-based RED-BIYO is proposed for detecting coconut tree damages using multimodal data namely stem image and audio signal. The stem images are gathered by an image capture device and audio signals are captured by a coconut microphone sensor. The stem image is pre-processed by Savitzky-Golay (SavG) filter to remove the noise from the images. The YOLOv9 model leverages a ResNeSt-based split-attention backbone and a Feature Pyramid Network (FPN) to effectively extract detailed multi-scale features from pre-processed images for enabling precise detection of Damaged Tree (DT) and No Damaged Tree (NDT). Simultaneously, the acquired audio signals are pre-processed using short-time framing with Hamming windowing and STFT-based spectrogram extraction. The resulting spectrograms are analyzed using the Spiking Convolutional Neural Network based Bidirectional Gated Recurrent Unit (SCNN-BiGRU) for RPW detection. SCNN extracts spatial and frequency features, while BiGRU captures temporal patterns in both forward and backward directions. The multi model data are fused in fuzzy rules to assess coconut tree health and notify farmers via an IoT-enabled Blynk app. The proposed RED-BIYO achieves the detection accuracy (AC) of 99.54

This paper addresses the problem of simultaneous estimation of faults and attacks in networked Takagi–Sugeno fuzzy chaotic systems. To this end, an enhanced multiple intermediate estimator framework is developed to reconstruct both fault and attack signals in a unified manner without relying on restrictive observer matching conditions. To further improve the performance of the estimations, an optimization strategy is proposed based on the genetic algorithm to optimize the most important parameters of the estimators in a structured way. The tuning problem is outlined as an optimization problem that attempts to minimize the estimation errors of fault and attack signals. Using the global search feature of the genetic algorithm, the proposed approach does not require manual parameter selection and has a higher level of estimation accuracy. Based on the obtained estimates, a compensation mechanism is incorporated to mitigate the adverse effects of faults and attacks on system performance. Sufficient conditions ensuring the stability of the closed-loop system are derived using Lyapunov theory and expressed in terms of linear matrix inequalities. Finally, numerical simulations are provided to validate the effectiveness of the proposed approach. The results demonstrate that the genetic algorithm-optimized estimator significantly improves estimation accuracy and enhances the resilience of the system against simultaneous faults and cyber-attacks.

Landslides threaten human life, infrastructure, and environmental stability, necessitating rapid response systems enabled by accurate and timely detection. This study carried out a comparative evaluation of the deep learning methods against the traditional machine learning algorithms in the detection of landslides using satellite remote sensing imagery. The proposed CNN-based architecture, comprising VGG and ResNet models, is compared with conventional ML algorithms, namely, Random Forest, Support Vector Machine, Decision Tree, K-Nearest Neighbors, and Logistic Regression. Experiments were carried out on two benchmark datasets Recent Landslide Database and Relict Landslide Database using False Positive Rate (FPR) and Matthews Correlation Coefficient as key metrics. Results reveal that the proposed ResNet gives the minimum FPR (0.0683% and 0.1296% for RLD and LLD, respectively) and maximum MCC values (0.7012 for RLD and 0.724 for LLD), thus besting all traditional models. VGG gives competitively very high MCC scores with stable accuracy. RF offers good accuracy but suffers from more false positives, while SVM and KNN offer the worst of classification results, especially on LLD. Additionally, an Average Relative Predictor Importance (ARPI) study is performed that distinguishes slope gradient and curvature as the most important features for landslide prediction.

This paper proposes a novel terahertz biosensor employing a multi-layer architecture of graphene, MXene and silver to detect brain tumors with high sensitivity. The proposed structure comprises a silicon dioxide substrate with periodic rectangular resonator arrays which provide dual band operation at two frequency bands of 0.1-0.3 THz and 0.4-0.45 THz. Finite element simulations performed in COMSOL Multiphysics demonstrated that the absorption characteristics of the sensor are strongly modulated by both the graphene chemical potential and the electromagnetic wave angle of incidence; peak absorptances of 98.7% and 97.3% are achieved under optimal conditions for each parameter respectively, with all absorption values verified to satisfy energy conservation (R + T + A = 1) through cosine-corrected S-parameter normalization. The refractive index sensitivity of the biosensor was found to be exceptionally high at 1538 GHz/RIU over the refractive index range of 1.3333-1.4833 RIU, spanning the biologically validated contrast between healthy and tumor-affected brain tissue. A very strong linear relationship between the resonance frequency shift and the refractive index of the biosensor yields a calibration curve F = - 0.1550n + 0.6402 with a correlation coefficient of 99.326%. Machine learning validation with Locally Weighted Linear Regression (LWLR), conducted on a noise-augmented dataset of 980 samples with 10-fold cross-validation and independent test-set evaluation, shows that the proposed biosensor has high prediction accuracy with R2 values as high as 95%. These results demonstrate that the proposed biosensor has significant potential for early stage brain tumor diagnostics and supports an integrated sensor-to-inference pipeline suitable for future clinical deployment.

The systematic system that involves modern IoT and embedded system techniques to provide real-time awareness of accidents and rescue signals to road users. The system relies on current modules of fall detection sensor, GPSbased landmarking and GSM communication technologies to facilitate convenient emergency call-in networks to the stranded victims in emergency cases. Fall recognition is more accurate and efficient, with the inclusion of accelerator and gyroscopes sensors, combined with the location detection systems based on GPS, especially in numerous implementation conditions. This study spearheads the safety technology development with its inexpensive portable tool filling the emergency communication gaps to follow the provision of safe chances of road safety and personal surveillance. Accelerator and gyroscope sensors, which include GPS modules, GSM modules, and assistive technology implementations have been integrated into the research as the fundamental components of the research.

In real life, recognizing human Emotions is the important role because of reactions to different emotional expresses which affect brain signals. Many scientific researchers have been concerned with construction of an automatic system to recognize human emotions. Machines have the ability to recognize human emotions; they can actually look inside the user's mind and act based on the observed mental position. This is possible by the efficient algorithms used to extract features, building the models and the classification algorithms. Nowadays, speech Signals have more attention to this kind of research, though 35% of emotions can be realized from speech. This research mainly focuses on proposing an automated method for identifying emotions using CNN model based on the Speech signals. Using branched CNN and MFCC, features 85% accuracy has been achieved. The same models can also be applied to different emotional datasets. In addition to speech signals, face images are also involved to make sure that the predicted emotion is with maximum accuracy. For the emotion recognition using the face expressions, CNN model is developed and by usage of FER database, model is trained and tested with an accuracy of 68%. Hence it is perceived that emotion predicted from speech signals performs high when related to facial emotions. Higher accuracy can be acquired by the same system by fusing both facial emotions with speech in future.

Breast cancer remains the leading cause of cancer-related mortality among women worldwide, with survival rates declining sharply from over 90

Pediatric pneumonia is the leading cause of disease and death. Preventing complications and death requires early diagnosis. Due of their variety, standard chest radiograph interpretation requires robust deep learning models for accurate and automatic identification. This study uses transformer-augmented encoding (TAE) to improve pediatric chest X-ray features. The study uses global contextual reasoning and localized radiographic texture clues to classify pneumonia precisely. The hybrid architecture TAE-CXR uses convolutional layers for low-level spatial encoding and multi-scale transformer blocks for long-range interactions. Transformer-augmented attention modules and cross-feature fusion identify pneumonia-specific radiography patches and eliminate background noise. On a pediatric chest radiograph dataset, TAE-CXR outperforms CNN and pure ViT models in sensitivity, specificity, and accuracy. The framework displays pneumonia regions using attention maps to simplify clinical use. In conclusion, transformer-augmented encoding greatly enhances pneumonia identification in child radiography. Results indicate that this technology could be a reliable tool for pediatric imaging decision support.

This research presents an AI-powered assistive glove that uses the signals of EMG to interpret muscle activity for the support of patients with motor impairments. Real-time machine learning classifiers analyze such signals for the correct recognition and activation of hand gestures. This low-cost wearable can go a long way in giving better outcomes to rehabilitation and helping patients become independent while performing basic tasks. By integrating EMG signal acquisition with machine-learning-based classification and a lightweight, simplified actuation mechanism, the proposed glove assists users in recovering partial hand functionality and enhances their ability to perform everyday activities. The system incorporates a personalized calibration framework, enabling the glove to adapt to individual variations in muscle strength and movement patterns, thereby improving usability and performance. Adaptive Machine Learning improved gesture recognition accuracy over repeated and guided hand movements. The glove also provides real-time feedback, which helps users to relearn motor functions through hand movements. The design focus on energy efficiency, ensuring battery life for continuous daily usage. It is low cost and also portable which makes it suitable for rehabilitation centers and for personal use compared to the bulky and costly robotic systems. All in all, the proposed glove is optimized and enhances independence, supports therapy progress and contributes to accessible assistive technology solutions.

This paper develops, tests, and evaluates neural network-based plant disease classification methods using healthy, powdery, and rust datasets. Agriculture requires an early plant disease diagnosis to avoid crop losses. Given their picture categorization abilities, Convolutional Neural Networks (CNNs) can diagnose plant diseases. In this study, healthy and diseased plant leaves were photographed. Labelling images by class facilitates supervised neural network learning. Resizing, normalization, and supplementation improve dataset quality and diversity. Using CNN architecture for classification. After optimizing model parameters with labeled training data, the learning algorithm tests them on unseen test data during training, validation, and testing. Trial results contain accuracy curves showing training epoch performance and model accuracy on the test set as a percentage. The model displays its class classification performance in a confusion matrix. Critical data analysis proves the tactics work and suggests future effort. Key findings emphasize the importance of neural network-based plant disease detection systems and their potential impact on farming.

Automation of facial emotion recognition is an important branch of artificial intelligence and computer vision that has many potential applications in mental health diagnostics, human–computer interaction and security. The existing methods, however, usually have weaknesses in robustness, scalability and computational efficiency. This work proposes a self-attention-based Vision Transformer method that treats images as sequences of patches to capture global dependencies and spatial relations more effectively than other methods. The model is trained and evaluated using a large-scale dataset. On average, the model achieves an overall accuracy of 97

This work introduces a high-performance graphene-silver hybrid metasurface biosensor for the fast and precise detection of COVID-19. Through parametric optimization with COMSOL Multiphysics, the sensor achieves a sensitivity of 400 GHz/RIU, a figure of merit (FOM) of 5.000 RIU⁻¹, and a Q factor of 12.7 within the refractive index range of 1.334-1.355 RIU. A machine learning framework enhances predictive reliability across different refractive indices, as reflected by a coefficient of determination (R²) of 0.90. The fabrication strategy-combining CVD graphene growth, electron beam lithography, and silver deposition-ensures scalability and practical realization. The novelty of this study lies in the synergistic integration of a graphene-silver metasurface platform with machine learning-based predictive modeling, enabling rapid, label-free, and highly accurate COVID-19 detection. Unlike conventional RT-PCR and antigen-based tests, which suffer from delays, high costs, or reduced sensitivity in asymptomatic cases, the proposed sensor achieves superior balance between sensitivity, figure of merit, and predictive accuracy, thereby surpassing state-of-the-art optical and terahertz biosensors. This positions the device as a novel, portable, and cost-effective diagnostic tool for next-generation pandemic preparedness.

Magnetic Resonance Imaging (MRI) classification and segmentation of brain tumors are still an important problem in medical imaging because tumors are typically heterogeneous and multicellular. Old-fashioned diagnostics do not scale, generalize or compute efficiently, and hence do not apply clinically. In order to overcome these problems, the research introduces a new Dynamic Attention-Augmented Neural Network (DAANN) that will improve the classification accuracy, robustness and computational efficiency of brain tumors. The proposed DAANN features a dynamic attention system that enables it to choose the diagnostically relevant areas from MRI images automatically, enabling it to perform even in noisy or sparse data. Combined with feature extraction over time, the model captures morphological and sequential dependencies to render full tumor analysis. When trained on benchmark MRI images, the DAANN had a classification precision of 94.8% and topped models like CNN and BiLSTM. The model was also robust to noise - with 89.0% accuracy under moderate noise levels - and operated efficiently using very little training data, with 78.0% accuracy on 10% of the training set. These findings point to the DAANN's real-time clinical availability in resource-constrained settings. As it takes up the challenge of what is possible, this work adds value to medical imaging as it provides a scalable and reproducible way to diagnose brain tumors. The work in the future will be to extend the model to multi-organ classification and to make it more easily read by clinical decision makers.

The performance of Mobile Ad-Hoc Networks (MANETs) can be verified based on number of input packets and life time of protocol requirements. Distance based geographic routing protocol are used to provide secure routing under congestion control and identifies packet loss. This paper presents performance evaluation and comparison of MANETs by using congestion on limited buffer requirements and malicious packet loss route discovery. We proposed distance based geographic routing protocol to monitor the routing requests, polling response and reduce the routing delays. In this paper analyze path control, available resources, and conditional delays and simulate the networks. This paper uses to reveal both buffer size and network performance characteristics. Analyze the exact frame work of throughput capacity, packet loss ratio and end-to-end delay in MANETs. In this mechanism reduce the overhead of routing strategies and calculate estimated distance over propagation of each node. Simulation experiments are verified by various expected values based on changes in number of hops and routing paths. In the protocol, every node evaluates the link quality and eliminates weak node or malicious node in MANET. Based on ready queue random early detection methods are used to reduce the network traffic delays and propagation delays. So this paper clearly indicates the MANETs performance under congestion control, packet loss and gives clear details about the life time of the networks and packet delay.

The increasing demand of integration density improvement and battery-powered device efficiency reduced complementary metal-oxide semiconductor (CMOS) technology node. In the technology of CMOS, the components are mainly affected with leakage power, dynamic switching power, short circuit power, gate oxide tunneling leakage current, sub threshold leakage current, and so on. To reduce the above limitations, design of high efficient low power static logic circuit using shorted-gate (SG) fin field-effect transistor (FinFET) based INput DEPendent (INDEP) in 22 nm CMOS technology (SLC-SG-FinFET-INDEP-22 nm CMOS) approach is proposed in this manuscript. The better selection of inputs to proposed INDEP FinFETs model is used for reducing leakage power. The efficiency of the proposed SLC-SG-FinFET-INDEP-22 nm CMOS technique is analysed using delay, power, power delay product, and stability analysis using noise margin. Thus, the proposed SLC-SG-FinFET-INDEP-22 nm CMOS has attained 21.31%, 41.47% and 12.7% lower delay, 20.87%, 34.5% and 22.41% lower power and 4.5%, 25.7% and 32.11% higher speed than existing methods static logic circuit input-controlled leakage restrainer transistor in 22 nm CMOS (SLC-ICLRT-22 nm CMOS), static logic circuit using self-control leakage-suppression block in 22 nm CMOS technology (SLC-SCLSB-22 nm CMOS), and static logic circuit using computational digital low dropout in 22 nm CMOS technology (SLC-CDLDO-22 nm CMOS) methods, respectively.

Magnetic resonance imaging (MRI) is a powerful tool for tumor diagnosis in human brain. Here, the MRI images are considered to detect the brain tumor and classify the regions as meningioma, glioma, pituitary and normal types. Numerous existing methods regarding brain tumor detection were suggested previously, but none of the methods accurately categorizes the brain tumor and consumes more computation period. To address these problems, an Evolutionary Gravitational Neocognitron Neural Network optimized with Marine Predators Algorithm is proposed in this article for MRI Brain Tumor Classification (EGNNN-VGG16-MPA-MRI-BTC). Initially, the brain MRI pictures are collected under Brats MRI image dataset. By using Savitzky-Golay Denoising approach, these images are pre-processed. The features are extracted utilizing visual geometry group network (VGG16). By utilizing VGG16, the features, like Grey level features, Haralick Texture features are extracted. These extracted features are given to EGNNN classifier, which categorizes the brain tumor as glioma, meningioma, pituitary gland and normal. Batch Normalization (BN) layer of EGNNN is eliminated and included with VGG16 layer. Marine Predators Optimization Algorithm (MPA) optimizes the weight parameters of EGNNN. The simulation is activated in MATLAB. Finally, the EGNNN-VGG16-MPA-MRI-BTC method attains 38.98%, 46.74%, 23.27% higher accuracy, 24.24%, 37.82%, 13.92% higher precision, 26.94%, 47.04%, 38.94% higher sensitivity compared with the existing AlexNet-SVM-MRI-BTC, RESNET-SGD-MRI-BTC and MobileNet-V2-MRI-BTC models respectively.