Pub Date : 2025-10-09DOI: 10.1109/TCE.2025.3619781
Ishtiaq Ahmad;Umair Ahmad Mughal;Liang Yang;Yazeed Alkhrijah;Ahmad Almadhor;Mohamad A. Alawad;Chau Yuen
The rapid advancement of consumer connected and autonomous vehicle (CAV) technologies offers significant improvements in transportation efficiency, safety, and user convenience. However, these benefits come with substantial cybersecurity risks, as in-vehicle networks and cloud connectivity expose CAVs to increasingly sophisticated cyberattacks. Conventional intrusion detection systems (IDS) often fall short in this domain, as they are not adaptive and struggle to handle the dynamic and stealthy nature of modern attacks. To address these limitations, we propose a novel IDS framework based on a stacking ensemble architecture that integrates multiple machine learning algorithms, Random Forest (RF), Support Vector Machine (SVM), Adaptive Boosting (AdaBoost), and Extreme Gradient Boosting (XGBoost), as base learners. A Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) serves as the meta-learner to capture temporal dependencies and sequential patterns in network traffic. To enhance the model’s generalization capability, we incorporate a model-agnostic meta-learning (MAML) approach into the LSTM-RNN meta-learner. The MAML-enhanced set of capabilities enables more effective detection of evolving and previously unseen attack scenarios. Simulation results demonstrate that the proposed framework consistently outperforms standalone LSTM-RNN models, traditional ensemble methods, and individual base learners in detecting complex cyberattack patterns in consumer CAV environments. These findings highlight the potential of meta-learning-driven ensemble IDS frameworks for securing next-generation intelligent transportation systems.
{"title":"Intelligent Ensemble Learning Framework for Intrusion Detection in Consumer Connected and Autonomous Vehicles","authors":"Ishtiaq Ahmad;Umair Ahmad Mughal;Liang Yang;Yazeed Alkhrijah;Ahmad Almadhor;Mohamad A. Alawad;Chau Yuen","doi":"10.1109/TCE.2025.3619781","DOIUrl":"https://doi.org/10.1109/TCE.2025.3619781","url":null,"abstract":"The rapid advancement of consumer connected and autonomous vehicle (CAV) technologies offers significant improvements in transportation efficiency, safety, and user convenience. However, these benefits come with substantial cybersecurity risks, as in-vehicle networks and cloud connectivity expose CAVs to increasingly sophisticated cyberattacks. Conventional intrusion detection systems (IDS) often fall short in this domain, as they are not adaptive and struggle to handle the dynamic and stealthy nature of modern attacks. To address these limitations, we propose a novel IDS framework based on a stacking ensemble architecture that integrates multiple machine learning algorithms, Random Forest (RF), Support Vector Machine (SVM), Adaptive Boosting (AdaBoost), and Extreme Gradient Boosting (XGBoost), as base learners. A Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) serves as the meta-learner to capture temporal dependencies and sequential patterns in network traffic. To enhance the model’s generalization capability, we incorporate a model-agnostic meta-learning (MAML) approach into the LSTM-RNN meta-learner. The MAML-enhanced set of capabilities enables more effective detection of evolving and previously unseen attack scenarios. Simulation results demonstrate that the proposed framework consistently outperforms standalone LSTM-RNN models, traditional ensemble methods, and individual base learners in detecting complex cyberattack patterns in consumer CAV environments. These findings highlight the potential of meta-learning-driven ensemble IDS frameworks for securing next-generation intelligent transportation systems.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 4","pages":"12437-12448"},"PeriodicalIF":10.9,"publicationDate":"2025-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778147","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-10-06DOI: 10.1109/TCE.2025.3618175
Muhammad Shafiq;Penghui Li;Lihua Yin;Nada Alasbali;Mohammad Mahtab Alam
The rapid proliferation of consumer IoT applications has embedded interconnected devices into daily life, creating highly dynamic and heterogeneous environments that pose significant security challenges. Diverse devices, protocols, and user-driven interactions complicate cyber threat analysis, while existing deep learning methods struggle with single-view models that fail to capture comprehensive behavioral patterns and multi-view approaches that suffer from ineffective feature fusion, leading to low generalization in dynamic scenarios. To address the problem, a novel technique named GRIOT-FENCE is first proposed. Then, based on GRIOT-FENCE, a new algorithm named DYNAMO-IoT is developed and designed. GRIOT-FENCE conducts comprehensive cyber threat analysis by modeling structural interactions, temporal dynamics, and statistical characteristics of network traffic across diverse consumer IoT devices, enhancing data security through robust threat detection. Its context-aware fusion module, FUSCONET, dynamically weights predictions based on device roles and network conditions, improving threat analysis transparency by highlighting critical behavioral features. The DYNAMO-IoT algorithm continuously monitors performance and triggers lightweight retraining of the fusion layer, ensuring adaptability to evolving cyber threats with minimal computational overhead, suitable for resource-constrained consumer environments. Experimental results demonstrate that GRIOT-FENCE achieves superior detection accuracy and robust threat analysis compared to state-of-the-art methods on benchmark IoT datasets, safeguarding consumer IoT applications and enhancing their trustworthiness through improved data security, system reliability, and transparent threat insights.
{"title":"GRIOT-FENCE: Multi-View Adaptive Intrusion Detection for Trustworthy Consumer IoT Cyber Threat Analysis","authors":"Muhammad Shafiq;Penghui Li;Lihua Yin;Nada Alasbali;Mohammad Mahtab Alam","doi":"10.1109/TCE.2025.3618175","DOIUrl":"https://doi.org/10.1109/TCE.2025.3618175","url":null,"abstract":"The rapid proliferation of consumer IoT applications has embedded interconnected devices into daily life, creating highly dynamic and heterogeneous environments that pose significant security challenges. Diverse devices, protocols, and user-driven interactions complicate cyber threat analysis, while existing deep learning methods struggle with single-view models that fail to capture comprehensive behavioral patterns and multi-view approaches that suffer from ineffective feature fusion, leading to low generalization in dynamic scenarios. To address the problem, a novel technique named GRIOT-FENCE is first proposed. Then, based on GRIOT-FENCE, a new algorithm named DYNAMO-IoT is developed and designed. GRIOT-FENCE conducts comprehensive cyber threat analysis by modeling structural interactions, temporal dynamics, and statistical characteristics of network traffic across diverse consumer IoT devices, enhancing data security through robust threat detection. Its context-aware fusion module, FUSCONET, dynamically weights predictions based on device roles and network conditions, improving threat analysis transparency by highlighting critical behavioral features. The DYNAMO-IoT algorithm continuously monitors performance and triggers lightweight retraining of the fusion layer, ensuring adaptability to evolving cyber threats with minimal computational overhead, suitable for resource-constrained consumer environments. Experimental results demonstrate that GRIOT-FENCE achieves superior detection accuracy and robust threat analysis compared to state-of-the-art methods on benchmark IoT datasets, safeguarding consumer IoT applications and enhancing their trustworthiness through improved data security, system reliability, and transparent threat insights.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 4","pages":"12449-12463"},"PeriodicalIF":10.9,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778139","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-10-06DOI: 10.1109/TCE.2025.3618484
Fangyu Liu;Hao Wang;Fangmin Sun;Xiangchen Li;Ye Li
Accurate objective physical fatigue (OPF) assessment is critical for enhancing safety in labor-intensive industries, optimizing athletic performance, and supporting personalized health management. With the advancement of wearable sensing technologies, wearable OPF assessment methods based on multi-source information fusion is emerging is gradually becoming an critical approach for sports and health monitoring. However, wearable OPF assessment methods face the following challenges: spatial modeling deficiency, ineffective multimodal fusion and limited long-range dependency capture. To address these limitations, we propose DBMNet, a dual-branch multi-modal network with spatial-temporal fusion for fatigue level classification. DBMNet captures temporal dynamics and spatial patterns separately from raw 12-lead electrocardiogram (ECG) and multi-site inertial measurement unit (IMU) signals. A novel Convolutional Additive Adaptive Cross-refinement Fusion (CAACF) module is introduced to enable adaptive, efficient, and interpretable fusion of heterogeneous modalities. We collected synchronized ECG and IMU data from 65 subjects during treadmill exercise following a modified Bruce protocol, annotated with both coarse- and fine-grained fatigue levels. Extensive experiments demonstrate that DBMNet achieves significant improvements over baseline and state-of-the-art methods, with up to 7% increase in classification accuracy and superior performance across multiple evaluation metrics. Moreover, DBMNet maintains a lightweight architecture suitable for deployment on mobile and wearable devices. This work provides an effective and scalable framework for real-time, objective fatigue monitoring using multi-source physiological signals.
{"title":"DBMNet: Dual-Branch Multi-Modal Network With Spatial-Temporal Fusion for Objective Physical Fatigue Assessment","authors":"Fangyu Liu;Hao Wang;Fangmin Sun;Xiangchen Li;Ye Li","doi":"10.1109/TCE.2025.3618484","DOIUrl":"https://doi.org/10.1109/TCE.2025.3618484","url":null,"abstract":"Accurate objective physical fatigue (OPF) assessment is critical for enhancing safety in labor-intensive industries, optimizing athletic performance, and supporting personalized health management. With the advancement of wearable sensing technologies, wearable OPF assessment methods based on multi-source information fusion is emerging is gradually becoming an critical approach for sports and health monitoring. However, wearable OPF assessment methods face the following challenges: spatial modeling deficiency, ineffective multimodal fusion and limited long-range dependency capture. To address these limitations, we propose DBMNet, a dual-branch multi-modal network with spatial-temporal fusion for fatigue level classification. DBMNet captures temporal dynamics and spatial patterns separately from raw 12-lead electrocardiogram (ECG) and multi-site inertial measurement unit (IMU) signals. A novel Convolutional Additive Adaptive Cross-refinement Fusion (CAACF) module is introduced to enable adaptive, efficient, and interpretable fusion of heterogeneous modalities. We collected synchronized ECG and IMU data from 65 subjects during treadmill exercise following a modified Bruce protocol, annotated with both coarse- and fine-grained fatigue levels. Extensive experiments demonstrate that DBMNet achieves significant improvements over baseline and state-of-the-art methods, with up to 7% increase in classification accuracy and superior performance across multiple evaluation metrics. Moreover, DBMNet maintains a lightweight architecture suitable for deployment on mobile and wearable devices. This work provides an effective and scalable framework for real-time, objective fatigue monitoring using multi-source physiological signals.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 4","pages":"12527-12538"},"PeriodicalIF":10.9,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778126","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-10-06DOI: 10.1109/TCE.2025.3618544
Faeiz Alserhani;Kamran Ahmad Awan;Amjad Alsirhani;Nabil Almashfi;Korhan Cengiz
Secure communication in resource-constrained Internet of Things (IoT) deployments—such as consumer electronics and smart city infrastructures—faces persistent challenges from both limited processing capacity and increasingly adaptive cyber threats. This work introduces the Adversarial Neural Encryption Framework (ANEF), which models encryption as a non-invertible mapping trained through an adversarial three-agent setup comprising an encoder (Alice), a decoder (Bob), and an adversary (Eve). The framework integrates entropy-regularized encryption, dynamic key scheduling with periodic reseeding, and quantization-aware optimization to support 8-bit inference on constrained hardware. Adversarial resistance is reinforced through residual masking and curvature-based penalties, improving robustness against adaptive attacks. Training is guided by a hybrid objective that balances reconstruction fidelity, orthogonality preservation, and adversarial minimization, applied over sequential encryption states. Experiments using the UNSW-NB15 dataset evaluate accuracy, communication overhead, and adversary success rate across diverse packet sizes. ANEF achieves 94.7% decryption accuracy with a 10.3% transmission overhead, while maintaining a 4.1% adversary success rate.
{"title":"ANEF: Adversarial Neural Encryption Framework for Secured Consumer Electronics in Smart Cities","authors":"Faeiz Alserhani;Kamran Ahmad Awan;Amjad Alsirhani;Nabil Almashfi;Korhan Cengiz","doi":"10.1109/TCE.2025.3618544","DOIUrl":"https://doi.org/10.1109/TCE.2025.3618544","url":null,"abstract":"Secure communication in resource-constrained Internet of Things (IoT) deployments—such as consumer electronics and smart city infrastructures—faces persistent challenges from both limited processing capacity and increasingly adaptive cyber threats. This work introduces the Adversarial Neural Encryption Framework (ANEF), which models encryption as a non-invertible mapping trained through an adversarial three-agent setup comprising an encoder (Alice), a decoder (Bob), and an adversary (Eve). The framework integrates entropy-regularized encryption, dynamic key scheduling with periodic reseeding, and quantization-aware optimization to support 8-bit inference on constrained hardware. Adversarial resistance is reinforced through residual masking and curvature-based penalties, improving robustness against adaptive attacks. Training is guided by a hybrid objective that balances reconstruction fidelity, orthogonality preservation, and adversarial minimization, applied over sequential encryption states. Experiments using the UNSW-NB15 dataset evaluate accuracy, communication overhead, and adversary success rate across diverse packet sizes. ANEF achieves 94.7% decryption accuracy with a 10.3% transmission overhead, while maintaining a 4.1% adversary success rate.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 4","pages":"12429-12436"},"PeriodicalIF":10.9,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778223","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-09-19DOI: 10.1109/TCE.2025.3611978
Yongsheng Du;Hongwei Sun
Traffic congestion is one of the biggest challenges faced by modern cities, causing delays, increasing fuel consumption, and contributing to air pollution. Cities worldwide are struggling to implement efficient traffic management systems that can adapt in real time to changing road conditions. Traditional traffic control methods, such as fixed-timing signals and centralized adaptive systems, fail to handle dynamic congestion efficiently which led to high communication costs, latency issues and privacy concerns. To address these issues this study introduces Federated Learning-Based Dynamic Traffic Management (FL-DTM), a decentralized traffic optimization framework designed for Internet of Things (IoT) enabled smart cities. Instead of sending raw data to a central server, traffic nodes such as traffic lights, connected vehicles, and roadside units (RSUs) train local machine learning models using real-time traffic flow data. These local models share only their learned parameters with a global server, which then aggregates updates and redistributes an improved global model back to local nodes. This ensures that traffic control decisions are made efficiently while maintaining data privacy and reducing communication overhead. The study uses SUMO to test the effectiveness of FL-DTM in different traffic conditions. Performance is compared against traditional fixed-timing traffic signals (FTTS) and centralized AI-based adaptive systems (CAIS). Results show that FL-DTM reduces average vehicle delay by 38.5%. It improves overall traffic throughput by 24.7%, and decreases fuel consumption by 16.3%. Additionally, model training time is reduced by 41% due to decentralized learning which make real-time adaptation faster. The incorporation of FL-DTM in traffic management enhances privacy, scalability, and computational efficiency.
{"title":"Federated-Learning-Based Dynamic Traffic Management for IoT-Enabled Smart Cities","authors":"Yongsheng Du;Hongwei Sun","doi":"10.1109/TCE.2025.3611978","DOIUrl":"https://doi.org/10.1109/TCE.2025.3611978","url":null,"abstract":"Traffic congestion is one of the biggest challenges faced by modern cities, causing delays, increasing fuel consumption, and contributing to air pollution. Cities worldwide are struggling to implement efficient traffic management systems that can adapt in real time to changing road conditions. Traditional traffic control methods, such as fixed-timing signals and centralized adaptive systems, fail to handle dynamic congestion efficiently which led to high communication costs, latency issues and privacy concerns. To address these issues this study introduces Federated Learning-Based Dynamic Traffic Management (FL-DTM), a decentralized traffic optimization framework designed for Internet of Things (IoT) enabled smart cities. Instead of sending raw data to a central server, traffic nodes such as traffic lights, connected vehicles, and roadside units (RSUs) train local machine learning models using real-time traffic flow data. These local models share only their learned parameters with a global server, which then aggregates updates and redistributes an improved global model back to local nodes. This ensures that traffic control decisions are made efficiently while maintaining data privacy and reducing communication overhead. The study uses SUMO to test the effectiveness of FL-DTM in different traffic conditions. Performance is compared against traditional fixed-timing traffic signals (FTTS) and centralized AI-based adaptive systems (CAIS). Results show that FL-DTM reduces average vehicle delay by 38.5%. It improves overall traffic throughput by 24.7%, and decreases fuel consumption by 16.3%. Additionally, model training time is reduced by 41% due to decentralized learning which make real-time adaptation faster. The incorporation of FL-DTM in traffic management enhances privacy, scalability, and computational efficiency.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 4","pages":"12332-12344"},"PeriodicalIF":10.9,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778184","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-09-17DOI: 10.1109/TCE.2025.3607447
Chengyuan Yan;Jianwei Xia;Jing Zhang;Ju H. Park;Hao Shen
This article addresses adaptive fixed-time event-triggered asymptotic tracking control problem for nonlinear switched systems under arbitrary switching. By adding exponential terms with factor to fixed-time controller and using a modified fixed-time lemma, a novel fixed-time stability criterion is given. Subsequently, to remove the singularity problem and achieve asymptotic tracking control, some piecewise functions and positive integral time-varying functions are proposed in the adaptive backstepping process. In addition, the mode-dependent dynamic event-triggered strategy of subsystem is designed to solve asynchronous switching problem between subsystems and the corresponding controller, which removes the restriction on the number of switch. Then, by means of common Lyapunov function method and Lyapunov stability theory, it is proved that all signals are bounded. In the end, the consumer robotic manipulator system is used to demonstrate the effectiveness of the proposed control algorithm.
{"title":"Fuzzy Logic-Based Fixed-Time Tracking Control for Nonlinear Switched Systems: An Event-Driven Approach","authors":"Chengyuan Yan;Jianwei Xia;Jing Zhang;Ju H. Park;Hao Shen","doi":"10.1109/TCE.2025.3607447","DOIUrl":"https://doi.org/10.1109/TCE.2025.3607447","url":null,"abstract":"This article addresses adaptive fixed-time event-triggered asymptotic tracking control problem for nonlinear switched systems under arbitrary switching. By adding exponential terms with factor to fixed-time controller and using a modified fixed-time lemma, a novel fixed-time stability criterion is given. Subsequently, to remove the singularity problem and achieve asymptotic tracking control, some piecewise functions and positive integral time-varying functions are proposed in the adaptive backstepping process. In addition, the mode-dependent dynamic event-triggered strategy of subsystem is designed to solve asynchronous switching problem between subsystems and the corresponding controller, which removes the restriction on the number of switch. Then, by means of common Lyapunov function method and Lyapunov stability theory, it is proved that all signals are bounded. In the end, the consumer robotic manipulator system is used to demonstrate the effectiveness of the proposed control algorithm.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 4","pages":"12418-12428"},"PeriodicalIF":10.9,"publicationDate":"2025-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778150","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 increasing complexity of consumer electronics and their dependence on continuous multivariate sensor data make intelligent anomaly detection a critical task. This paper proposes a novel anomaly detection framework, termed FD-Diffusion (Feature-Driven Diffusion Framework), tailored to time-series data generated by consumer electronic devices. The framework is composed of three key components. First, the Status-Aware Multimodal Encoder (SAME) extracts temporally aligned and semantically enriched representations from heterogeneous sensor streams by incorporating device status, behavioral context, and operational states. Second, the Condition-Aware Diffusion Model with Residual Refinement (CADM) leverages these embeddings to guide the reverse denoising trajectory of a diffusion process, ensuring that reconstructed sequences conform to normal behavior patterns. To enhance sensitivity to subtle deviations, CADM includes a residual refinement decoder with localized attention. Third, the Multi-Scale Adaptive Anomaly Scoring Mechanism (MAASM) fuses reconstruction loss, latent divergence, and semantic inconsistency into an interpretable and adaptive anomaly score, adjusted dynamically based on recent operational statistics. Experimental results on real-world device datasets demonstrate that FD-Diffusion outperforms existing methods in both detection accuracy and interpretability, offering a scalable solution for intelligent monitoring of consumer electronics under dynamic and diverse usage conditions.
{"title":"Intelligent Anomaly Detection Method for Consumer Electronics Based on Feature-Driven Learning and Diffusion Model","authors":"Guodong Wang;Qianqian Li;Qun Wang;Hadeel Alsolai;Xuejia Jiang","doi":"10.1109/TCE.2025.3608449","DOIUrl":"https://doi.org/10.1109/TCE.2025.3608449","url":null,"abstract":"The increasing complexity of consumer electronics and their dependence on continuous multivariate sensor data make intelligent anomaly detection a critical task. This paper proposes a novel anomaly detection framework, termed FD-Diffusion (Feature-Driven Diffusion Framework), tailored to time-series data generated by consumer electronic devices. The framework is composed of three key components. First, the Status-Aware Multimodal Encoder (SAME) extracts temporally aligned and semantically enriched representations from heterogeneous sensor streams by incorporating device status, behavioral context, and operational states. Second, the Condition-Aware Diffusion Model with Residual Refinement (CADM) leverages these embeddings to guide the reverse denoising trajectory of a diffusion process, ensuring that reconstructed sequences conform to normal behavior patterns. To enhance sensitivity to subtle deviations, CADM includes a residual refinement decoder with localized attention. Third, the Multi-Scale Adaptive Anomaly Scoring Mechanism (MAASM) fuses reconstruction loss, latent divergence, and semantic inconsistency into an interpretable and adaptive anomaly score, adjusted dynamically based on recent operational statistics. Experimental results on real-world device datasets demonstrate that FD-Diffusion outperforms existing methods in both detection accuracy and interpretability, offering a scalable solution for intelligent monitoring of consumer electronics under dynamic and diverse usage conditions.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 4","pages":"12500-12509"},"PeriodicalIF":10.9,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778136","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Increasing urbanization and the global transition toward sustainable, eco-friendly energy systems require efficient and robust energy predictions for smart grids. The inherently unpredictable, volatile, and intermittent nature of energy demand necessitates an accurate short-term load forecasting model to ensure reliable consumer applications. However, conventional deep learning models often struggle to address complex and dynamic load patterns. To address these challenges, this research presents a novel trustworthy GAI-assisted model comprising i) a variational autoencoder that maps raw energy consumption data to extract meaningful and compact features and ii) a deep learning model utilizing a dual attention mechanism with convolutional long short-term memory (DAConvLSTM), that effectively captures the temporal dependencies of the complex load pattern and optimizes forecasting accuracy. The effectiveness and robustness of the proposed model are extensively evaluated using publicly available comprehensive datasets. The results demonstrate the performance of the proposed model, with an overall improvement of 1.45%~81.54% in the mean absolute error, 1.92%~78.61% in the root mean square error, and 1.55%~81.85% in the mean absolute percentage error compared with other baseline methods. The results validate the effectiveness of the proposed model in predicting peak load demand and have practical implications, thereby enhancing the existing knowledge for creating robust energy management in smart grid applications.
{"title":"Trustworthy Load Forecasting With Generative AI: A Dual-Attention ConvLSTM and VAE-Based Approach","authors":"Abid Ali;Yuanqing Xia;Muhammad Fahad Zia;Waqas Haider Khan Bangyal;Muddesar Iqbal","doi":"10.1109/TCE.2025.3606753","DOIUrl":"https://doi.org/10.1109/TCE.2025.3606753","url":null,"abstract":"Increasing urbanization and the global transition toward sustainable, eco-friendly energy systems require efficient and robust energy predictions for smart grids. The inherently unpredictable, volatile, and intermittent nature of energy demand necessitates an accurate short-term load forecasting model to ensure reliable consumer applications. However, conventional deep learning models often struggle to address complex and dynamic load patterns. To address these challenges, this research presents a novel trustworthy GAI-assisted model comprising i) a variational autoencoder that maps raw energy consumption data to extract meaningful and compact features and ii) a deep learning model utilizing a dual attention mechanism with convolutional long short-term memory (DAConvLSTM), that effectively captures the temporal dependencies of the complex load pattern and optimizes forecasting accuracy. The effectiveness and robustness of the proposed model are extensively evaluated using publicly available comprehensive datasets. The results demonstrate the performance of the proposed model, with an overall improvement of 1.45%~81.54% in the mean absolute error, 1.92%~78.61% in the root mean square error, and 1.55%~81.85% in the mean absolute percentage error compared with other baseline methods. The results validate the effectiveness of the proposed model in predicting peak load demand and have practical implications, thereby enhancing the existing knowledge for creating robust energy management in smart grid applications.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 4","pages":"12490-12499"},"PeriodicalIF":10.9,"publicationDate":"2025-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778135","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-09-08DOI: 10.1109/TCE.2025.3607501
Guozhi Yan;Zuoxiu Yang;Weizhen Han;Bingyi Liu;Kai Liu
This letter presents a novel Traffic Signal and Vehicle Joint Control (TSVJC) framework that integrates traffic signal regulation and vehicle speed control to improve traffic efficiency. In this framework, each intersection operates an independent signal control process managed by dedicated signal agents, while nearby vehicle agents dynamically adjust their speeds through coordinated optimization. To enable effective cooperation between these heterogeneous agents, we propose the Multi-Agent Cooperative Attention Optimization (MACAO) algorithm, which employs Graph Attention Networks (GAT) to generate attention weights from observed traffic states across intersections. Simulation results show that the proposed approach significantly reduces vehicle travel time and improves traffic efficiency compared to existing methods.
{"title":"A Multi-Agent Cooperative Attention Framework for Joint Control of Traffic Signal and Vehicles","authors":"Guozhi Yan;Zuoxiu Yang;Weizhen Han;Bingyi Liu;Kai Liu","doi":"10.1109/TCE.2025.3607501","DOIUrl":"https://doi.org/10.1109/TCE.2025.3607501","url":null,"abstract":"This letter presents a novel Traffic Signal and Vehicle Joint Control (TSVJC) framework that integrates traffic signal regulation and vehicle speed control to improve traffic efficiency. In this framework, each intersection operates an independent signal control process managed by dedicated signal agents, while nearby vehicle agents dynamically adjust their speeds through coordinated optimization. To enable effective cooperation between these heterogeneous agents, we propose the Multi-Agent Cooperative Attention Optimization (MACAO) algorithm, which employs Graph Attention Networks (GAT) to generate attention weights from observed traffic states across intersections. Simulation results show that the proposed approach significantly reduces vehicle travel time and improves traffic efficiency compared to existing methods.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 4","pages":"12539-12541"},"PeriodicalIF":10.9,"publicationDate":"2025-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778130","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-08-26DOI: 10.1109/TCE.2025.3602923
Yusen Wang;Xiaolong Xu;Ruoshui Wang;Muhammad Bilal;Wei Liu;Guangming Cui
Against the backdrop of global carbon neutrality and low-carbon agriculture, the urgency to promote low-carbon agricultural consumer electronics through the integration of sustainable computing is increasingly evident. Edge servers, with their high efficiency and low latency characteristics, have become a crucial component of sustainable computing. Using their local deployment and low-latency advantages, edge servers enable a real-time decision optimization system, optimize energy-efficient resource scheduling, reduce carbon emissions in the agricultural production process, and thereby facilitate low-carbon agriculture. However, for edge computing to deliver efficient, low-latency, and low-energy services, it must rely on the strategic allocation of edge servers. Suboptimal deployment strategies can result in elevated network delays, diminished service reliability, and higher levels of carbon output. The problem of identifying the most effective locations for deploying a limited number of edge servers, while addressing key performance concerns such as latency, reliability, and environmental impact under practical constraints, is commonly known as the $k$ ESP problem. Recent research has addressed issues such as high latency, low robustness, and carbon emission reduction in edge computing networks, but has yet to simultaneously reduce latency, improve robustness, and optimize computing resources while lowering carbon emissions. To tackle this challenge, we introduce the $k$ ESP-PSO approach, designed to mitigate high latency, enhance service reliability, and reduce carbon emissions by determining an efficient deployment strategy for edge servers. Specifically, $k$ ESP-PSO method incorporates a Particle Swarm Optimization (PSO) algorithm, which iteratively refines the location of edge servers based on the spatial distribution of base stations and mobile users across the target region. Through this mechanism, $k$ ESP-PSO is capable of theoretically deriving the most effective configuration of edge server placements. Extensive experiments on Melbourne and Shanghai Telecom data sets demonstrate that the proposed method significantly reduces carbon emissions compared to baseline approaches, while also optimizing computing resources and effectively supporting low-carbon agricultural consumer electronics.
{"title":"A Resource-Efficient Placement of Edge Servers for Green Agriculture Consumer Electronics","authors":"Yusen Wang;Xiaolong Xu;Ruoshui Wang;Muhammad Bilal;Wei Liu;Guangming Cui","doi":"10.1109/TCE.2025.3602923","DOIUrl":"https://doi.org/10.1109/TCE.2025.3602923","url":null,"abstract":"Against the backdrop of global carbon neutrality and low-carbon agriculture, the urgency to promote low-carbon agricultural consumer electronics through the integration of sustainable computing is increasingly evident. Edge servers, with their high efficiency and low latency characteristics, have become a crucial component of sustainable computing. Using their local deployment and low-latency advantages, edge servers enable a real-time decision optimization system, optimize energy-efficient resource scheduling, reduce carbon emissions in the agricultural production process, and thereby facilitate low-carbon agriculture. However, for edge computing to deliver efficient, low-latency, and low-energy services, it must rely on the strategic allocation of edge servers. Suboptimal deployment strategies can result in elevated network delays, diminished service reliability, and higher levels of carbon output. The problem of identifying the most effective locations for deploying a limited number of edge servers, while addressing key performance concerns such as latency, reliability, and environmental impact under practical constraints, is commonly known as the <inline-formula> <tex-math>$k$ </tex-math></inline-formula>ESP problem. Recent research has addressed issues such as high latency, low robustness, and carbon emission reduction in edge computing networks, but has yet to simultaneously reduce latency, improve robustness, and optimize computing resources while lowering carbon emissions. To tackle this challenge, we introduce the <inline-formula> <tex-math>$k$ </tex-math></inline-formula>ESP-PSO approach, designed to mitigate high latency, enhance service reliability, and reduce carbon emissions by determining an efficient deployment strategy for edge servers. Specifically, <inline-formula> <tex-math>$k$ </tex-math></inline-formula>ESP-PSO method incorporates a Particle Swarm Optimization (PSO) algorithm, which iteratively refines the location of edge servers based on the spatial distribution of base stations and mobile users across the target region. Through this mechanism, <inline-formula> <tex-math>$k$ </tex-math></inline-formula>ESP-PSO is capable of theoretically deriving the most effective configuration of edge server placements. Extensive experiments on Melbourne and Shanghai Telecom data sets demonstrate that the proposed method significantly reduces carbon emissions compared to baseline approaches, while also optimizing computing resources and effectively supporting low-carbon agricultural consumer electronics.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"71 4","pages":"12373-12385"},"PeriodicalIF":10.9,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778124","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：