Pub Date : 2026-02-06DOI: 10.1016/j.compeleceng.2026.111005
Hoangcong Le, Cheng-Kai Lu
Real-time human action recognition (HAR) plays a vital role in healthcare monitoring, particularly for elderly care and assistive environments. However, existing HAR systems often struggle with high computational demands, large model sizes, and vulnerability to background noise, limiting their use on edge devices in the Internet of Medical Things (IoMT) settings. This study proposes Temporal-SpatialCNN, a lightweight framework built on a novel CNNBlock architecture that integrates convolution, batch normalization, and Spatial-Dropout to enhance both efficiency and generalization. The study systematically analyzes various layer arrangements within CNNBlock and identifies an optimal configuration that improves recognition performance while maintaining minimal computational overhead. The model incorporates diverse skeletal input modalities – joint coordinates, joint collection distances, slow motion, and velocity – to capture enriched spatio-temporal features. Extensive experiments on five well-known benchmark datasets validate the effectiveness of the proposed approach, achieving state-of-the-art accuracy (up to 99.66% on the Florence-3D dataset) with an inference time of 9.6 ms. To demonstrate real-world applicability, a real-time Save Our Soul (SOS) system was implemented on standard hardware, capable of detecting emergency gestures such as calls for assistance, thereby highlighting the model’s practical potential in real-time, resource-constrained healthcare scenarios.
{"title":"A low-latency deep learning approach for human action recognition in medical internet of things applications","authors":"Hoangcong Le, Cheng-Kai Lu","doi":"10.1016/j.compeleceng.2026.111005","DOIUrl":"10.1016/j.compeleceng.2026.111005","url":null,"abstract":"<div><div>Real-time human action recognition (HAR) plays a vital role in healthcare monitoring, particularly for elderly care and assistive environments. However, existing HAR systems often struggle with high computational demands, large model sizes, and vulnerability to background noise, limiting their use on edge devices in the Internet of Medical Things (IoMT) settings. This study proposes Temporal-SpatialCNN, a lightweight framework built on a novel CNNBlock architecture that integrates convolution, batch normalization, and Spatial-Dropout to enhance both efficiency and generalization. The study systematically analyzes various layer arrangements within CNNBlock and identifies an optimal configuration that improves recognition performance while maintaining minimal computational overhead. The model incorporates diverse skeletal input modalities – joint coordinates, joint collection distances, slow motion, and velocity – to capture enriched spatio-temporal features. Extensive experiments on five well-known benchmark datasets validate the effectiveness of the proposed approach, achieving state-of-the-art accuracy (up to 99.66% on the Florence-3D dataset) with an inference time of 9.6 ms. To demonstrate real-world applicability, a real-time Save Our Soul (SOS) system was implemented on standard hardware, capable of detecting emergency gestures such as calls for assistance, thereby highlighting the model’s practical potential in real-time, resource-constrained healthcare scenarios.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"132 ","pages":"Article 111005"},"PeriodicalIF":4.9,"publicationDate":"2026-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174182","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-05DOI: 10.1016/j.compeleceng.2026.110992
Reny Jose, K. Balasubramanian
Classification of cardamom pods remains a challenging task due to changes in lighting, inconsistencies in texture, and spectral features that overlap with existing methods, which lack the attributes of robustness, interpretability, and real-time usability. This research addresses the aforementioned issues by introducing an advanced dual-stream deep learning model, Di-stream improved Divine Religions Planarian ConvNeXt (DDRPCNeXt), which combines deep spatial features with handcrafted descriptors for enhanced discrimination of spectral and textural features. Supported by the Improved Divine Religions Algorithm (IDRA) metaheuristic optimizer, the model gets to enjoy fast convergence and stable training. Gradient-weighted Class Activation Mapping (Grad-CAM) guarantees that predictions are transparent and interpretable by showing visual heatmaps. The system is made available via a Kotlin cross-platform mobile application, which allows for on-device real-time classification. When tested with a dataset consisting of 1000 high-resolution cardamom images, the framework demonstrated outstanding performance, recording 99.25% accuracy, 99.09% precision, and 99.15% recall and F1-score. Such results validate the suggested solution as a tool for automated agricultural quality control that is accurate, interpretable, and ready to be used in the field.
{"title":"Optimized deep learning with Grad-CAM for automated cardamom classification: A multispectral imaging approach for real-time mobile deployment","authors":"Reny Jose, K. Balasubramanian","doi":"10.1016/j.compeleceng.2026.110992","DOIUrl":"10.1016/j.compeleceng.2026.110992","url":null,"abstract":"<div><div>Classification of cardamom pods remains a challenging task due to changes in lighting, inconsistencies in texture, and spectral features that overlap with existing methods, which lack the attributes of robustness, interpretability, and real-time usability. This research addresses the aforementioned issues by introducing an advanced dual-stream deep learning model, Di-stream improved Divine Religions Planarian ConvNeXt (DDRPCNeXt), which combines deep spatial features with handcrafted descriptors for enhanced discrimination of spectral and textural features. Supported by the Improved Divine Religions Algorithm (IDRA) metaheuristic optimizer, the model gets to enjoy fast convergence and stable training. Gradient-weighted Class Activation Mapping (Grad-CAM) guarantees that predictions are transparent and interpretable by showing visual heatmaps. The system is made available via a Kotlin cross-platform mobile application, which allows for on-device real-time classification. When tested with a dataset consisting of 1000 high-resolution cardamom images, the framework demonstrated outstanding performance, recording 99.25% accuracy, 99.09% precision, and 99.15% recall and F1-score. Such results validate the suggested solution as a tool for automated agricultural quality control that is accurate, interpretable, and ready to be used in the field.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"132 ","pages":"Article 110992"},"PeriodicalIF":4.9,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174251","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The Internet of Vehicles (IoV) enables intelligent transportation systems in the efficient management of autonomous vehicles (AVs) by enabling real-time data exchange and message communication over the Internet. Rapid advances in autonomous systems, hardware sensors, software, and the integration of artificial intelligence have enabled the development of automated services and improved convenience for users. In the IoV, heavy traffic, connectivity issues, and frequent handoffs create significant challenges. Unmanned aerial vehicles (UAVs, sometimes called drones) offer a promising solution to these issues by helping to alleviate network congestion and improving overall network performance. Because UAVs are highly mobile devices, potential security breaches pose a significant challenge for communication between UAVs and autonomous vehicles. To address this, we propose a secure and efficient ultralightweight protocol that uses UAV technology in an IoV environment. Our protocol employs elliptic-curve cryptography and cryptographic operators such as exclusive-or operations and one-way hashing. A formal security analysis of the protocol using the Scyther simulation tool reveals that it is resilient against security attacks, while an informal security analysis shows that the protocol is secure against several known security and privacy threats. The computational and communication costs of the proposed protocol are lower than those of other existing protocols, while being efficient, outperforming other solutions in terms of security features.
{"title":"An ultralightweight and reliable authentication protocol for secure communication in UAV-assisted IoAV systems","authors":"Sanjeev Kumar , Mohd Shariq , Gopal Singh Rawat , Muhammad Shafiq , Khalid Alsubhi , Mehedi Masud , Hossam Meshref","doi":"10.1016/j.compeleceng.2026.110998","DOIUrl":"10.1016/j.compeleceng.2026.110998","url":null,"abstract":"<div><div>The Internet of Vehicles (IoV) enables intelligent transportation systems in the efficient management of autonomous vehicles (AVs) by enabling real-time data exchange and message communication over the Internet. Rapid advances in autonomous systems, hardware sensors, software, and the integration of artificial intelligence have enabled the development of automated services and improved convenience for users. In the IoV, heavy traffic, connectivity issues, and frequent handoffs create significant challenges. Unmanned aerial vehicles (UAVs, sometimes called drones) offer a promising solution to these issues by helping to alleviate network congestion and improving overall network performance. Because UAVs are highly mobile devices, potential security breaches pose a significant challenge for communication between UAVs and autonomous vehicles. To address this, we propose a secure and efficient ultralightweight protocol that uses UAV technology in an IoV environment. Our protocol employs elliptic-curve cryptography and cryptographic operators such as exclusive-or operations and one-way hashing. A formal security analysis of the protocol using the Scyther simulation tool reveals that it is resilient against security attacks, while an informal security analysis shows that the protocol is secure against several known security and privacy threats. The computational and communication costs of the proposed protocol are lower than those of other existing protocols, while being efficient, outperforming other solutions in terms of security features.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"132 ","pages":"Article 110998"},"PeriodicalIF":4.9,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174181","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-04DOI: 10.1016/j.compeleceng.2026.111022
Monika Srivastava , Gautam Kumar , Rishav Singh
In medical imaging, the segmentation of skin lesions plays a vital role in detecting and treating skin cancer. Deep learning demonstrates its efficacy in this process. However, it largely relies on extensive and well-annotated datasets that are often limited by healthcare agencies privacy restrictions and institutional data silos. Federated Learning (FL) emerged as a boon, enabling collaborative training without sharing data. Yet, in a real-world setting, healthcare bodies may possess various computational capacities that can affect the consistency of the FL framework, posing the requirement of generalising the model architecture. This study proposes a Decentralised Federated Learning (DFL) framework to improve model generalisation for Skin Lesion Segmentation (SkinSeg). It incorporates a novel Voting (V)-based client selection mechanism to identify the most suitable local model based on performance metrics and dataset size. The selected model is then subjected to Gradual Pruning (GP) via a modified Lottery Ticket Hypothesis (LTH) to reduce model complexity while preserving segmentation quality. The pruned model is then broadcast to all clients for further training. The VGPDFL-SkinSeg substantially improved over State-Of-The-Art FL frameworks on benchmark datasets HAM10K, ISIC-2016/17/18 and DermIs+DermQuest. It achieved a client-wise average Dice Coefficient (DSC) of 90.09%, 96.60% Accuracy, 82.45% meanIOU, 13.63% HD95 and 5.20% ASSD. Initially, each client starts with different segmentation models, reflecting practical diverse systems, and gradually converges towards homogeneity. The study shows that gradual pruning up to 40% yields better segmentation quality than fixed pruning at the beginning and is consistent with client scaling.
{"title":"VGPDFL-SkinSeg: Enhancing model generalisation with data diversity via voting-based client selection and gradual pruning for decentralised federated skin lesion segmentation","authors":"Monika Srivastava , Gautam Kumar , Rishav Singh","doi":"10.1016/j.compeleceng.2026.111022","DOIUrl":"10.1016/j.compeleceng.2026.111022","url":null,"abstract":"<div><div>In medical imaging, the segmentation of skin lesions plays a vital role in detecting and treating skin cancer. Deep learning demonstrates its efficacy in this process. However, it largely relies on extensive and well-annotated datasets that are often limited by healthcare agencies privacy restrictions and institutional data silos. Federated Learning (FL) emerged as a boon, enabling collaborative training without sharing data. Yet, in a real-world setting, healthcare bodies may possess various computational capacities that can affect the consistency of the FL framework, posing the requirement of generalising the model architecture. This study proposes a Decentralised Federated Learning (DFL) framework to improve model generalisation for Skin Lesion Segmentation (SkinSeg). It incorporates a novel Voting (V)-based client selection mechanism to identify the most suitable local model based on performance metrics and dataset size. The selected model is then subjected to Gradual Pruning (GP) via a modified Lottery Ticket Hypothesis (LTH) to reduce model complexity while preserving segmentation quality. The pruned model is then broadcast to all clients for further training. The VGPDFL-SkinSeg substantially improved over State-Of-The-Art FL frameworks on benchmark datasets HAM10K, ISIC-2016/17/18 and DermIs+DermQuest. It achieved a client-wise average Dice Coefficient (DSC) of 90.09%, 96.60% Accuracy, 82.45% meanIOU, 13.63% HD95 and 5.20% ASSD. Initially, each client starts with different segmentation models, reflecting practical diverse systems, and gradually converges towards homogeneity. The study shows that gradual pruning up to 40% yields better segmentation quality than fixed pruning at the beginning and is consistent with client scaling.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"132 ","pages":"Article 111022"},"PeriodicalIF":4.9,"publicationDate":"2026-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174250","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-03DOI: 10.1016/j.compeleceng.2026.110990
Mohammad Bdaqli, Saeed Meshgini, Reza Afrouzian
Motor imagery classification using electroencephalography (EEG) signals is a fundamental component of Brain-Computer Interface (BCI) systems. It enables individuals with physical disabilities to control robotic limbs and perform various movements. However, the inherently noisy nature of EEG signals poses significant challenges for their effective utilization in this domain. In this study, we propose a novel end-to-end deep learning model based on feature fusion of multiple deep learning blocks, including a Convolutional Neural Network (CNN), Temporal Convolutional Network (TCN), and Squeeze and Excitation (SE) attention mechanism, enabling the model to learn discriminative features for classifying raw motor imagery signals without any preprocessing. The proposed architecture employs novel feature fusion strategies to maximize classification performance and computational efficiency. The CNN extracts initial spatial features, the TCN captures temporal dependencies, and the SE attention mechanism emphasizes the most informative features from the CNN output. The model was evaluated on the BCI Competition IV 2a and 2b datasets. Training was conducted for 500 epochs (2a dataset) and 200 epochs (2b dataset), using only the first session of each subject for training and validation. The average classification accuracies on the completely isolated test sets (second session) were 78.12 % and 85.72 % for the 2a and 2b datasets, respectively. These results demonstrate that the proposed model effectively classifies multi-class motor imagery signals.
利用脑电图（EEG）信号进行运动图像分类是脑机接口（BCI）系统的基本组成部分。它使身体残疾的人能够控制机械肢体并进行各种运动。然而，脑电信号固有的噪声特性对其在该领域的有效利用提出了重大挑战。在这项研究中，我们提出了一种新的端到端深度学习模型，该模型基于多个深度学习模块的特征融合，包括卷积神经网络（CNN）、时间卷积网络（TCN）和挤压和激励（SE）注意机制，使模型能够在不进行任何预处理的情况下学习判别特征，用于对原始运动图像信号进行分类。该体系结构采用新颖的特征融合策略，最大限度地提高分类性能和计算效率。CNN提取初始空间特征，TCN捕获时间依赖性，SE注意机制强调CNN输出中信息量最大的特征。该模型在BCI Competition IV 2a和2b数据集上进行了评估。对500个epoch （2a数据集）和200个epoch （2b数据集）进行训练，仅使用每个主题的第一个会话进行训练和验证。对于2a和2b数据集，完全隔离测试集（第二次）的平均分类准确率分别为78.12%和85.72%。结果表明，该模型能有效地对多类运动图像信号进行分类。
{"title":"A novel EEG motor imagery classification model using feature fusion of temporal convolution and attention","authors":"Mohammad Bdaqli, Saeed Meshgini, Reza Afrouzian","doi":"10.1016/j.compeleceng.2026.110990","DOIUrl":"10.1016/j.compeleceng.2026.110990","url":null,"abstract":"<div><div>Motor imagery classification using electroencephalography (EEG) signals is a fundamental component of Brain-Computer Interface (BCI) systems. It enables individuals with physical disabilities to control robotic limbs and perform various movements. However, the inherently noisy nature of EEG signals poses significant challenges for their effective utilization in this domain. In this study, we propose a novel end-to-end deep learning model based on feature fusion of multiple deep learning blocks, including a Convolutional Neural Network (CNN), Temporal Convolutional Network (TCN), and Squeeze and Excitation (SE) attention mechanism, enabling the model to learn discriminative features for classifying raw motor imagery signals without any preprocessing. The proposed architecture employs novel feature fusion strategies to maximize classification performance and computational efficiency. The CNN extracts initial spatial features, the TCN captures temporal dependencies, and the SE attention mechanism emphasizes the most informative features from the CNN output. The model was evaluated on the BCI Competition IV 2a and 2b datasets. Training was conducted for 500 epochs (2a dataset) and 200 epochs (2b dataset), using only the first session of each subject for training and validation. The average classification accuracies on the completely isolated test sets (second session) were 78.12 % and 85.72 % for the 2a and 2b datasets, respectively. These results demonstrate that the proposed model effectively classifies multi-class motor imagery signals.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"132 ","pages":"Article 110990"},"PeriodicalIF":4.9,"publicationDate":"2026-02-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174183","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Multi-replica and multi-cloud public auditing (MRMC-PA) is a method used to ensure data availability and integrity by verifying multiple copies of data stored across multiple cloud environments. However, existing MRMC-PA schemes are vulnerable to quantum attacks and incur high computational and communication overhead due to their reliance on pairing-based cryptography (PBC). In addition, they provide limited support for dynamic data operations across all replicas and suffer from either the certificate management problem (CMP) or the key escrow problem (KEP). To address these limitations, this paper proposes a quantum-resistant, multi-replica, and multi-cloud certificateless public auditing scheme (QR-MRMC-CLPAS) based on lattice-based cryptography over module lattices instead of PBC. The security of QR-MRMC-CLPAS is proven under the Module Learning With Errors (M-LWE) and Module Small Integer Solution (M-SIS) assumptions. To support data dynamics, we introduce a dynamic replica version table that ensures both consistency and integrity of multiple replicas across multi-cloud environments. Furthermore, the use of certificateless cryptography eliminates CMP and KEP. Performance analysis and experimental results demonstrate that QR-MRMC-CLPAS achieves significantly higher computational and communication efficiency compared to existing MRMC-PA schemes while ensuring strong quantum resilience.
{"title":"QR-MRMC-CLPAS: Quantum-resistant multi-replica and multi-cloud certificateless public auditing scheme based on module lattices","authors":"Renuka Cheeturi , Syam Kumar Pasupuleti , Rashmi Ranjan Rout","doi":"10.1016/j.compeleceng.2026.111000","DOIUrl":"10.1016/j.compeleceng.2026.111000","url":null,"abstract":"<div><div>Multi-replica and multi-cloud public auditing (MRMC-PA) is a method used to ensure data availability and integrity by verifying multiple copies of data stored across multiple cloud environments. However, existing MRMC-PA schemes are vulnerable to quantum attacks and incur high computational and communication overhead due to their reliance on pairing-based cryptography (PBC). In addition, they provide limited support for dynamic data operations across all replicas and suffer from either the certificate management problem (CMP) or the key escrow problem (KEP). To address these limitations, this paper proposes a quantum-resistant, multi-replica, and multi-cloud certificateless public auditing scheme (QR-MRMC-CLPAS) based on lattice-based cryptography over module lattices instead of PBC. The security of QR-MRMC-CLPAS is proven under the Module Learning With Errors (M-LWE) and Module Small Integer Solution (M-SIS) assumptions. To support data dynamics, we introduce a dynamic replica version table that ensures both consistency and integrity of multiple replicas across multi-cloud environments. Furthermore, the use of certificateless cryptography eliminates CMP and KEP. Performance analysis and experimental results demonstrate that QR-MRMC-CLPAS achieves significantly higher computational and communication efficiency compared to existing MRMC-PA schemes while ensuring strong quantum resilience.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"132 ","pages":"Article 111000"},"PeriodicalIF":4.9,"publicationDate":"2026-02-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174184","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-03DOI: 10.1016/j.compeleceng.2026.111006
Shamsul Haq, Prabal Verma
Healthcare 5.0 is a transformative paradigm that revolutionizes healthcare delivery and improves patient outcomes through incorporating cutting-edge technologies. In this alignment, the paper describes the understanding of Healthcare 5.0 involving different emerging technologies and their roles in effective decision outcomes with proper examples. In correspondence to the significance of Healthcare 5.0, the paper is preceded by focusing on the importance of cloud and edge computing in such environments. It also covers different tools and techniques, analytical methods and advanced emerging analytical algorithms for disease management and treatment optimization. Consequently, it examines the applications of edge computing with emerging analytical technologies in healthcare, showcasing various use cases such as remote patient monitoring, personalized medicine, intelligent healthcare systems, and data-driven decision support resulting in improved patient care and operational efficiency. Subsequently, the statistical results with the systematic framework are performed on the basis of 563 papers published in reputed journals and organizations for the comprehensive analysis of existing technologies and to identify research solutions and challenges in the development of Smart Healthcare. Finally, we summarize our key findings and propose future directions for research and smart healthcare development.
{"title":"An extensive examination of adaptive intelligence in cloud-to-edge systems for Healthcare 5.0","authors":"Shamsul Haq, Prabal Verma","doi":"10.1016/j.compeleceng.2026.111006","DOIUrl":"10.1016/j.compeleceng.2026.111006","url":null,"abstract":"<div><div>Healthcare 5.0 is a transformative paradigm that revolutionizes healthcare delivery and improves patient outcomes through incorporating cutting-edge technologies. In this alignment, the paper describes the understanding of Healthcare 5.0 involving different emerging technologies and their roles in effective decision outcomes with proper examples. In correspondence to the significance of Healthcare 5.0, the paper is preceded by focusing on the importance of cloud and edge computing in such environments. It also covers different tools and techniques, analytical methods and advanced emerging analytical algorithms for disease management and treatment optimization. Consequently, it examines the applications of edge computing with emerging analytical technologies in healthcare, showcasing various use cases such as remote patient monitoring, personalized medicine, intelligent healthcare systems, and data-driven decision support resulting in improved patient care and operational efficiency. Subsequently, the statistical results with the systematic framework are performed on the basis of 563 papers published in reputed journals and organizations for the comprehensive analysis of existing technologies and to identify research solutions and challenges in the development of Smart Healthcare. Finally, we summarize our key findings and propose future directions for research and smart healthcare development.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"132 ","pages":"Article 111006"},"PeriodicalIF":4.9,"publicationDate":"2026-02-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174186","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-02DOI: 10.1016/j.compeleceng.2026.111003
Arash Heidari , Ahmad Khonsari , Seyed Hamed Rastegar
Connected vehicles rely on continuous Vehicle-to-Everything (V2X) communication, which exposes the Internet of Vehicles (IoV) to latency-sensitive and privacy-critical cyberattacks. This paper presents Federated Learning with Intelligent Traffic-aware Energy optimization (FLITE), an energy-efficient, privacy-preserving framework for intrusion detection that trains a lightweight Gated Recurrent Unit (GRU) detector on vehicles using federated learning while keeping raw telemetry local. A deep reinforcement learning–based scheduler at roadside units selects clients and transmit powers based on data quality, channel state, and device energy, reducing redundant communication. Experiments on multiple vehicular and network intrusion datasets show that FLITE achieves up to 99.8% accuracy and improves F1-score and recall by about 2–3 percentage points over strong baselines, while reducing energy consumption by 36–45%, communication overhead by more than 60%, and detection delay by up to 60%. These results demonstrate that FLITE enables real-time, fleet-wide intrusion detection for large-scale IoV deployments under realistic resource constraints.
{"title":"An energy-efficient privacy-preserving framework for intrusion detection in the internet of vehicles","authors":"Arash Heidari , Ahmad Khonsari , Seyed Hamed Rastegar","doi":"10.1016/j.compeleceng.2026.111003","DOIUrl":"10.1016/j.compeleceng.2026.111003","url":null,"abstract":"<div><div>Connected vehicles rely on continuous Vehicle-to-Everything (V2X) communication, which exposes the Internet of Vehicles (IoV) to latency-sensitive and privacy-critical cyberattacks. This paper presents Federated Learning with Intelligent Traffic-aware Energy optimization (FLITE), an energy-efficient, privacy-preserving framework for intrusion detection that trains a lightweight Gated Recurrent Unit (GRU) detector on vehicles using federated learning while keeping raw telemetry local. A deep reinforcement learning–based scheduler at roadside units selects clients and transmit powers based on data quality, channel state, and device energy, reducing redundant communication. Experiments on multiple vehicular and network intrusion datasets show that FLITE achieves up to 99.8% accuracy and improves F1-score and recall by about 2–3 percentage points over strong baselines, while reducing energy consumption by 36–45%, communication overhead by more than 60%, and detection delay by up to 60%. These results demonstrate that FLITE enables real-time, fleet-wide intrusion detection for large-scale IoV deployments under realistic resource constraints.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"132 ","pages":"Article 111003"},"PeriodicalIF":4.9,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174202","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-02DOI: 10.1016/j.compeleceng.2026.111001
Dam Minh Linh , Huynh Trong Thua , Tran Cong Hung
Phishing attacks that evade traditional detection mechanisms by exploiting deceptive uniform resource locators (URLs) remain a significant cybersecurity threat. This study proposes an adaptive phishing URL detection framework that integrates Levenshtein distance-based string similarity, a hybrid embedding transformer (HET) encoder-based server-side verification mechanism, and a dynamically updated local blacklist. First, a rapid local lookup is executed to identify known phishing URLs. If the input URL is absent from the blacklist, the Levenshtein distance algorithm detects subtle character-level variations, identifying typosquatting and obfuscation effectively. For ambiguous cases, the HET-based module uses a lightweight post-hoc inference method that classifies URL embeddings via k-nearest neighbor voting based on Euclidean similarity in the latent space, thereby avoiding retraining and enabling real-time adaptation to emerging phishing threats. Confirmed phishing URLs are added iteratively to the local repository to improve detection continuously, enhancing future classification accuracy. Experimental evaluation on a large-scale dataset comprising 235,795 URLs revealed that the proposed method outperforms state-of-the-art approaches, achieving a detection accuracy of 99.8 %, with a false-positive rate of 0.441 % and false-negative rate of 0.0617 %. Additionally, real-time validation using a Chrome browser extension confirmed rapid processing, with an average processing time of 4.43–6.84 ms per URL on a dataset comprising 5,000 URLs. These results highlight the efficiency of the proposed framework in real-world cybersecurity contexts, enabling high detection accuracy, fast response times, and adaptability to evolving phishing strategies, and underscore the importance of proactive threat intelligence and real-time phishing mitigation in developing scalable, high-performance security infrastructures.
{"title":"Real-time phishing uniform resource locator detection based on hybrid embedding transformer and retraining-free inferencing","authors":"Dam Minh Linh , Huynh Trong Thua , Tran Cong Hung","doi":"10.1016/j.compeleceng.2026.111001","DOIUrl":"10.1016/j.compeleceng.2026.111001","url":null,"abstract":"<div><div>Phishing attacks that evade traditional detection mechanisms by exploiting deceptive uniform resource locators (URLs) remain a significant cybersecurity threat. This study proposes an adaptive phishing URL detection framework that integrates Levenshtein distance-based string similarity, a hybrid embedding transformer (HET) encoder-based server-side verification mechanism, and a dynamically updated local blacklist. First, a rapid local lookup is executed to identify known phishing URLs. If the input URL is absent from the blacklist, the Levenshtein distance algorithm detects subtle character-level variations, identifying typosquatting and obfuscation effectively. For ambiguous cases, the HET-based module uses a lightweight post-hoc inference method that classifies URL embeddings via k-nearest neighbor voting based on Euclidean similarity in the latent space, thereby avoiding retraining and enabling real-time adaptation to emerging phishing threats. Confirmed phishing URLs are added iteratively to the local repository to improve detection continuously, enhancing future classification accuracy. Experimental evaluation on a large-scale dataset comprising 235,795 URLs revealed that the proposed method outperforms state-of-the-art approaches, achieving a detection accuracy of 99.8 %, with a false-positive rate of 0.441 % and false-negative rate of 0.0617 %. Additionally, real-time validation using a Chrome browser extension confirmed rapid processing, with an average processing time of 4.43–6.84 ms per URL on a dataset comprising 5,000 URLs. These results highlight the efficiency of the proposed framework in real-world cybersecurity contexts, enabling high detection accuracy, fast response times, and adaptability to evolving phishing strategies, and underscore the importance of proactive threat intelligence and real-time phishing mitigation in developing scalable, high-performance security infrastructures.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"132 ","pages":"Article 111001"},"PeriodicalIF":4.9,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174203","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01DOI: 10.1016/j.compeleceng.2026.110997
Jaber Parchami , Seyed Reza Talebiyan , Abbas Abdulhussein Dahham , Dhulfiqar Dhurgham Husam , Ali Darroudi
The increasing prevalence of cyber threats has made malware detection a critical task for ensuring digital security. In this study, we propose a novel hybrid approach, termed Hybrid Deep Learning Network with Multilevel Feature Selection (HDLNet-MFS), for the classification and detection of various types of malwares. The proposed HDLNet-MFS framework employs a two-stage architecture comprising feature extraction and feature selection. To extract discriminative features from the two-dimensional representations of malware samples, a parallel combination of the pre-trained Inception V3 network and the Gray Level Co-occurrence Matrix (GLCM) algorithm is utilized, enabling the simultaneous capture of spatial and statistical texture features. For feature selection, a hybrid Neighborhood Component Analysis (NCA) - Minimum Redundancy Maximum Relevance (mRMR) algorithm is introduced, which effectively reduces feature dimensionality and ranks features based on their relevance to the classification task, allowing for the selection of the most informative attributes. Moreover, to address the data imbalance problem, especially for minority classes, Generative Adversarial Networks (GANs) are employed to augment the training data. The proposed approach aims to tackle key challenges such as class imbalance, limited training samples, high-dimensional feature spaces, and redundancy, thereby offering a robust and efficient solution for accurate malware classification. Experimental results demonstrate that HDLNet-MFS achieves an average classification accuracy of 99.74 % across 25 malware classes on the MalImg dataset, highlighting the precision, robustness, and effectiveness of the proposed system. Furthermore, the model exhibits high computational efficiency, achieving an average inference time of 0.84 seconds, which underscores its suitability for real-time or near–real-time malware detection scenarios in practical cybersecurity environments. The complete implementation of the proposed method is publicly available at: github.com/jaberparchami-tech/Malware-Detection-Hybrid-Framework.
{"title":"A hybrid deep learning approach for malware detection using generative adversarial network-based augmentation and multilevel feature selection","authors":"Jaber Parchami , Seyed Reza Talebiyan , Abbas Abdulhussein Dahham , Dhulfiqar Dhurgham Husam , Ali Darroudi","doi":"10.1016/j.compeleceng.2026.110997","DOIUrl":"10.1016/j.compeleceng.2026.110997","url":null,"abstract":"<div><div>The increasing prevalence of cyber threats has made malware detection a critical task for ensuring digital security. In this study, we propose a novel hybrid approach, termed Hybrid Deep Learning Network with Multilevel Feature Selection (HDLNet-MFS), for the classification and detection of various types of malwares. The proposed HDLNet-MFS framework employs a two-stage architecture comprising feature extraction and feature selection. To extract discriminative features from the two-dimensional representations of malware samples, a parallel combination of the pre-trained Inception V3 network and the Gray Level Co-occurrence Matrix (GLCM) algorithm is utilized, enabling the simultaneous capture of spatial and statistical texture features. For feature selection, a hybrid Neighborhood Component Analysis (NCA) - Minimum Redundancy Maximum Relevance (mRMR) algorithm is introduced, which effectively reduces feature dimensionality and ranks features based on their relevance to the classification task, allowing for the selection of the most informative attributes. Moreover, to address the data imbalance problem, especially for minority classes, Generative Adversarial Networks (GANs) are employed to augment the training data. The proposed approach aims to tackle key challenges such as class imbalance, limited training samples, high-dimensional feature spaces, and redundancy, thereby offering a robust and efficient solution for accurate malware classification. Experimental results demonstrate that HDLNet-MFS achieves an average classification accuracy of 99.74 % across 25 malware classes on the MalImg dataset, highlighting the precision, robustness, and effectiveness of the proposed system. Furthermore, the model exhibits high computational efficiency, achieving an average inference time of 0.84 seconds, which underscores its suitability for real-time or near–real-time malware detection scenarios in practical cybersecurity environments. The complete implementation of the proposed method is publicly available at: github.com/jaberparchami-tech/Malware-Detection-Hybrid-Framework.</div></div>","PeriodicalId":50630,"journal":{"name":"Computers & Electrical Engineering","volume":"132 ","pages":"Article 110997"},"PeriodicalIF":4.9,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174249","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号