IEEE Transactions on Consumer Electronics最新文献

{"title":"A Suspicious Anchor Detection Framework to Enhance Adversarial WSN Localization for Consumer Applications","authors":"Mazhar Islam;Muhammad Ikram;Khursheed Aurangzeb;Yu Xue","doi":"10.1109/TCE.2025.3625151","DOIUrl":"https://doi.org/10.1109/TCE.2025.3625151","url":null,"abstract":"Accurate localization is a critical challenge in wireless sensor networks (WSNs) deployed in smart consumer environments such as homes, precision agriculture, wearable healthcare systems, and connected appliances. These systems are particularly vulnerable in the presence of suspicious anchor nodes in an adversarial environment. Suspicious anchors may deliberately manipulate the measured data, leading to significant degradation in localization accuracy. Detecting and isolating these anchors remains a challenging problem. This paper presents a novel two-stage scheme to detect the suspicious anchor node and improve localization accuracy. The approach integrates an artificial intelligence technique, such as an enhanced Isolation Forest, with Sequential Probability Ratio Test (SPRT) based decision-making. The first stage enhances the conventional Isolation Forest algorithm to efficiently isolate suspicious anchor nodes based on multilateration distance data. We define a novel anomaly score function by integrating the traditional score of an anomaly with the average variation degree to select the reference anchors. In the next stage, the algorithm models a reference error interval using the Received Signal Strength Indicator (RSSI) and Time Difference of Arrival (TDOA) distance measurements and employs an improved SPRT mechanism to verify the suspicious anchors statistically. Extensive simulations were conducted to demonstrate the superiority of the proposed algorithm over conventional secure localization schemes. Simulation results show that the proposed algorithm significantly improves the attacker anchor detection. It achieves approximately 97% detection accuracy at <inline-formula> <tex-math>$mu _{delta } =3 times 10^{-8}s$ </tex-math></inline-formula> and reaches 100% detection at higher attack intensities while reducing false alarm rates to less than 8%. The algorithm significantly outperforms conventional schemes while preserving the integrity of localization services in consumer environments.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"7212-7225"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148837296","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}

{"title":"Explainable AI Framework for Retinal Images Collected by Consumer Electronics: Research on Early Screening of Uveitis and Adherence Improvement in Connected Healthcare Patient Follow-Up","authors":"Lili Hu;Yuxuan Peng;Hang Zhang;Jing Wang;Yue Bin;Rui Chang","doi":"10.1109/TCE.2026.3676209","DOIUrl":"https://doi.org/10.1109/TCE.2026.3676209","url":null,"abstract":"This article presents an explainable artificial intelligence (XAI) framework to assist with the detection of posterior segment inflammatory signs associated with uveitis through retinal images obtained via consumer electronics. We investigate the integration of deep learning architectures with interpretability mechanisms to enhance diagnostic transparency while assessing the framework’s impact on patient appointment adherence within connected healthcare ecosystems. Our proposed Uveitis-XAI-Net architecture employs a multi-scale convolutional neural network backbone integrated with dual explainability modules utilizing Gradient-weighted Class Activation Mapping (Grad-CAM) and Layer-wise Relevance Propagation (LRP). The framework was validated using a multi-center dataset comprising 4826 fundus images captured via smartphone-based imaging devices from 1543 participants across diverse clinical settings. The explainable AI system was deployed within a connected healthcare platform delivering personalized follow-up appointment reminders coupled with AI-generated diagnostic reports. Experimental evaluation demonstrates that the proposed framework achieves an area under the receiver operating characteristic curve (AUC-ROC) of 0.924, sensitivity of 87.3%, and specificity of 89.6% for detecting uveitis-related posterior segment findings, with explainable outputs exhibiting strong concordance with expert ophthalmologist annotations (Cohen’s Kappa <inline-formula> <tex-math>$kappa =0.81$ </tex-math></inline-formula>). A prospective 12-month clinical evaluation involving 798 patients revealed that XAI-enhanced digital health intervention significantly improved follow-up appointment adherence rates compared to standard care (78.5% vs. 52.4%, <inline-formula> <tex-math>$p$ </tex-math></inline-formula> < 0.001) and disease knowledge scores (mean improvement 24.7 percentage points, <inline-formula> <tex-math>$p$ </tex-math></inline-formula> < 0.001). These findings substantiate that incorporating explainable AI into consumer-grade imaging equipment supports detection of uveitis-associated retinal manifestations while enhancing patient engagement and long-term follow-up maintenance through transparent diagnostic assessments.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"6964-6975"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148837364","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}

{"title":"GCMT-Net: Generative Context-Driven Multimodal Temporal Network for Video Anomaly Detection in Consumer Applications","authors":"Liangbin Gao;Minmin Cheng;Kai Cui;Hongwei Lu;Wei Feng;Shenghao Liu;Xianjun Deng","doi":"10.1109/TCE.2026.3679820","DOIUrl":"https://doi.org/10.1109/TCE.2026.3679820","url":null,"abstract":"Video constitutes a significant content modality within consumer electronics applications. Video Anomaly Detection (VAD) aims to automatically identify anomalous events deviating from normal patterns in consumer videos. This technology directly supports quality assurance and user experience optimization in consumer electronics products, while also enabling functionalities such as environmental safety monitoring for consumer electronics applications. However, traditional methods rely heavily on single-modal features, making it difficult to effectively utilize the complementary clues between different modalities. They have limitations and semantic gaps in explaining complex scenes and understanding the video context. Moreover, most methods overlook complex cross-modal correlations and fail to fully exploit the potential of multimodal semantic clues. To address these challenges, this work proposes an innovative Generative Context-driven Multimodal Temporal Network (GCMT-Net) to detect abnormal video for consumer electronic applications. The GCMT-Net introduces a context generation network that transforms visual context into rich textual descriptions, alleviating semantic gaps that are difficult to capture with traditional visual features. A novel multi-scale temporal module then integrates visual, auditory, and textual information, significantly enhancing the model’s ability to recognize complex anomalous events and improving its overall robustness. Furthermore, modal consistency learning ensures semantic alignment between modalities and consistent feature representations across scales. Experiments on four benchmark datasets demonstrate the superior performance of the proposed framework in addressing VAD challenges in complex consumer application scenarios.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"5579-5593"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148837373","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}

{"title":"MMFashion+: Multimodal Federated Learning for Personalized Clothing Recommendation","authors":"Xinhui Yu;Peizhi Yan;Sophie Liu;Chunhua Wu","doi":"10.1109/TCE.2026.3674794","DOIUrl":"https://doi.org/10.1109/TCE.2026.3674794","url":null,"abstract":"Federated clothing recommendation aims to suggest clothing items to customers (users) based on their past purchase behaviors in a privacy-preserving manner, enhancing user experience while safeguarding privacy. Current federated recommendation systems face two main challenges. The first is limited information exploration. Existing methods mainly rely on item ID-based embeddings and ignore the assistance of multimodal information about items. The second challenge is data heterogeneity, caused by users perceiving the same clothing item differently, which leads to performance degradation. We therefore propose a multimodal federated learning framework for personalized clothing recommendation (MMFashion+). To address the first challenge, we design two branches to capture user preferences for clothing design and price, respectively. The design preference branch leverages image and text modalities, while the price preference branch incorporates item price information. To better model user design preferences, we introduce a threshold-based regularization loss to push the learned embeddings of user preferences and dislikes apart. To address the second challenge, we introduce personalized image and text adapters to guide the local model in capturing item attributes that align with users’ interests. This approach enables user-specific item representations that effectively capture individual design preferences. To evaluate the effectiveness of MMFashion+, we construct four datasets by extracting four clothing types (i.e., long sleeves, pants, dresses, and outerwear) from a real-world clothing dataset released by a fashion retail company. Experimental results demonstrate that MMFashion+ consistently outperforms existing methods, confirming its effectiveness in addressing the two key challenges in federated clothing recommendation.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"5544-5557"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836460","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}

{"title":"HATSAS: Hybrid Attack Tolerant Watermarking Framework for Speech and Audio Signals","authors":"Subreena Mushtaq;Samrah Mehraj;Shabir A. Parah","doi":"10.1109/TCE.2026.3656957","DOIUrl":"https://doi.org/10.1109/TCE.2026.3656957","url":null,"abstract":"With the exponential growth of digital audio transmission through consumer devices like smartphones, smart speakers, iPads, digital audio players etc., copyright protection has become critical for consumer electronics ecosystem. To address this challenge, this study presents a new hybrid attack-tolerant audio watermarking framework, called HATSAS, for copyright protection of consumer multimedia applications. A novel technique called Neighborhood Correlation Adjustment (NCA) has been proposed, and used in HATSAS along with fast Walsh-Hadamard transform (FWHT) to significantly improve the robustness to singular, and hybrid attacks. HATSAS exploits the correlation properties of audio segments for watermark embedding and comprises two key stages: structured segmentation, and watermark embedding employing neighborhood-based block selection called NCA strategy. HATSAS achieves high imperceptibility, with a Signal-to-Noise Ratio (SNR) between 36 to 42 dB thereby, maintaining audio quality standards expected by consumers. HATSAS achieves BER reduction by 61% and 11.5% under Gaussian noise (20 & 15 dB), 38% reduction under echo addition over the best existing method, and supports a payload of 204.8 bps, demonstrating strong resilience to various distortions. Therefore, HATSAS is well-suited for integration into consumer electronics devices, ensuring robust protection of multimedia content in real-world consumer environments.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"5307-5316"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836448","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}

{"title":"Robust and Adaptive Compliant Deep Reinforcement Learning With Vision-Force Integration for Embodied Multiple Peg-in-Hole Assembly","authors":"Le Fu;Yaolin Tang;Xiaoge Xue;Qi Fang;Huaguang Shi;Xibeng Zhang;Lei Shi;Yi Zhou","doi":"10.1109/TCE.2026.3675608","DOIUrl":"https://doi.org/10.1109/TCE.2026.3675608","url":null,"abstract":"Robotic multiple peg-in-hole assembly is a typical contact-rich manipulation task in embodied intelligence, with applications from housework to consumer electronics production. However, this task involves handling strong geometric coupling, in which existing methods lack sufficient compliance and adaptability to pose deviations and environmental uncertainty, thereby increasing the risk of jamming and task failure. To address these issues, this paper proposes a framework based on deep reinforcement learning with vision-force integration. A refined Data-regularized Q module enhances visual feature robustness under target pose uncertainty by using synchronized augmentation for stacked frames with a temporal-consistency regularization. Meanwhile, a hybrid reward mechanism accelerates policy convergence by adaptively weighting external task rewards and intrinsic curiosity rewards to balance exploration and exploitation, providing denser learning signals and preventing premature convergence to local optima. Variable impedance control restores compliance through real-time stiffness adjustment for safe and compliant physical interaction. Experiments on diverse electrical socket configurations demonstrate reliable sim-to-real transfer, robust performance under challenging perturbations, and moderated contact forces, highlighting the practical potential of the proposed approach for autonomous robotic assembly.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"7540-7551"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836560","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}

{"title":"DSTEGNN: A Dynamic Spatio-Temporal Explainer Graph Neural Network for Securing Consumer Devices Against Malware Attacks","authors":"Le Sun;Yuyan Guo;Ghulam Muhammad","doi":"10.1109/TCE.2026.3678908","DOIUrl":"https://doi.org/10.1109/TCE.2026.3678908","url":null,"abstract":"The rapid proliferation of consumer devices and mobile applications brings convenience but also amplifies malware-driven security and privacy risks. In particular, AI-enabled malware attacks pose severe threats to both device security and user privacy by leveraging AI to evade traditional defenses. Although deep learning techniques have improved malware detection, their decisions remain opaque and highly sensitive to data quality, limiting their trustworthiness in real-world deployment. To address these limitations, we present the Dynamic Spatio-Temporal Explainer Graph Neural Network (DSTEGNN), an explainable malware detection framework tailored for consumer applications. DSTEGNN incorporates an adaptive edge learning mechanism that automatically infers communication dependencies from traffic features, enabling the graph structure to evolve with dynamic network behaviors rather than relying on manually predefined edges. In addition, we develop Global-GNNExplainer, a global-level interpretability module that extracts dataset-wide critical substructures, revealing consistent behavior patterns associated with malicious activities and offering stable, semantically meaningful explanations beyond instance-level analysis. By integrating spatio-temporal attention with graph convolution, DSTEGNN achieves high detection accuracy while maintaining interpretability. Experiments on public datasets demonstrate that DSTEGNN not only improves malware detection performance but also enhances resilience and trustworthiness in defending consumer devices against AI-enabled attacks.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"7374-7384"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836703","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}

{"title":"Generalized Multimodal Brain–Computer Interfaces: From Neural Signal Sensing to Scenario-Adaptive Decoding","authors":"Shiqiang Ma;Tengfei Xiao;Yanyan Shen;Yang Zhan;Baiying Lei;Michael Kwok-Po Ng;Shuqiang Wang","doi":"10.1109/TCE.2026.3677888","DOIUrl":"https://doi.org/10.1109/TCE.2026.3677888","url":null,"abstract":"Fueled by concurrent advances in sensing and computing, brain–computer interfaces are evolving from single-electroencephalography modalities toward generalized fusion systems that synergistically integrate electrical, hemodynamic, optical, magnetic, and peripheral physiological signals. Flexible dry-electrode arrays and optically pumped magnetometers now deliver high spatio-temporal resolution within everyday wearable form factors; cross-modal attention and graph-network approaches unify millisecond-scale electrophysiology with second-scale hemodynamics at the feature level; decision-level and deep-learning frameworks further enable adaptive modality weighting and robust transfer. Building on these advances, we propose the concept of “generalized fusion,” emphasizing dynamic synergy among heterogeneous signals across the entire encoding–decoding–feedback loop, and discuss the emerging role of generative AI in completing missing modalities, balancing signal-to-noise trade-offs, and expanding demographic coverage. Finally, we summarize the distinct demands of medical rehabilitation, human–computer interaction, and intelligent control scenarios on fusion accuracy and user load, highlighting that hardware comfort, algorithmic real-time capability, and multidimensional evaluation must be continually refined to translate multimodal brain–computer interfaces from laboratory validation into real-world deployment.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"6120-6131"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836696","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}

{"title":"Multimodal Semantic-Similarity Gate for Real-Time Optimized Deepfake Detection","authors":"Yousung Yeon;JunHo Yoon;Chang Choi","doi":"10.1109/TCE.2026.3666953","DOIUrl":"https://doi.org/10.1109/TCE.2026.3666953","url":null,"abstract":"With recent advances in deep learning-based generative artificial intelligence (AI), highly realistic synthetic videos and audio have rapidly proliferated and been increasingly misused, intensifying the need for reliable deepfake detection. Accordingly, growing attention has been given to multimodal AI-based deepfake detection methods. These systems typically analyze the global context of video and audio information using attention mechanisms to detect whether the content is synthetic. However, attention-based multimodal fusion does not distinguish the manipulation status of each modality; therefore, the computational cost increases as the video and audio sequences become longer. Although frame selection can reduce computation, it often causes global structural information loss and omits certain important sequence information. This study proposes a multimodal semantic-similarity gate (MSG)-based deepfake detection framework, which reduces computational cost while preserving the entire video–audio sequence information. The MSG adopts a two-stage backbone to process common sequences of video and audio modalities, comprising a front stage that determines the semantic similarity between modalities and a back stage that controls hard/soft attention. Experimental results show that the proposed MSG maintains an F1-score difference within 0.0001–0.0106 compared with baseline models, while reducing GFLOPs by 10.59–28.07% and per-sample inference latency by 3.35–41.96%. These results demonstrate improved computational efficiency and support real-time deepfake detection.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"7160-7168"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836744","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}

{"title":"Combining Lightweight and Post-Quantum Cryptography With Physically Unclonable Functions for Secure In-Vehicle Communications","authors":"Arcangelo Castiglione;Teresa Elia;Christian Esposito;Franco Cirillo","doi":"10.1109/TCE.2026.3684325","DOIUrl":"https://doi.org/10.1109/TCE.2026.3684325","url":null,"abstract":"Modern vehicles are complex cyber-physical systems that rely on In-Vehicle networks to coordinate the operation of multiple Electronic Control Units (ECUs). Among the various communication protocols used in this domain, the Controller Area Network (CAN) remains the most widely adopted due to its reliability and performance. However, it is inherently insecure, making vehicles increasingly vulnerable to cyber threats, especially in the context of growing connectivity and emerging post-quantum risks. This paper proposes a protocol to enhance the security of CAN networks, combining Physically Unclonable Functions (PUFs) for efficient ECU authentication, Lightweight AEAD (Authenticated Encryption with Associated Data), and Post-Quantum Cryptography for the confidentiality of communications. The protocol stands out for its low computational overhead, requiring only <inline-formula> <tex-math>$2n$ </tex-math></inline-formula> messages to authenticate all <inline-formula> <tex-math>$n$ </tex-math></inline-formula> nodes in the network, and for its compatibility with the standard 8-byte CAN frame. This aspect makes our protocol a powerful, robust, high-performance, and scalable security solution that can be integrated into current and next-generation in-vehicle networks, making it suitable for the safety-critical automotive context.","PeriodicalId":13208,"journal":{"name":"IEEE Transactions on Consumer Electronics","volume":"72 3","pages":"7169-7184"},"PeriodicalIF":9.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836745","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}