Internet of Things最新文献_第10页

CONTEXT-NET: A context-aware nexus-based aggregation protocol for opportunistic networks 上下文- net：机会网络的上下文感知的基于网络的聚合协议

IF 7.6 3区计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Internet of Things

Pub Date : 2025-10-20 DOI: 10.1016/j.iot.2025.101809

Rounak Raman , Ayush Yadav , Deepika Kukreja , Deepak Kumar Sharma

Opportunistic Networks enable communication in dynamic, resource-constrained environments using a store-carry-forward approach. However, challenges such as efficient data aggregation, collision avoidance, minimizing data redundancy, and trust management persist. This study proposes the Context-Aware Nexus-Based Aggregation Protocol (CONTEXT-NET), which integrates spatial, temporal, and contextual dimensions for optimized data transmission. CONTEXT-NET employs a nexus ring topology with synchronized sector-based scheduling, autoencoder-based dimensionality reduction, and a hybridized Ant Colony Optimization (ACO)-like routing algorithm for adaptive routing, ensuring minimal collisions and efficient data aggregation. A trust-based scoring system enhances security by identifying and excluding unreliable nodes. The dataset for analysis consists of a customized random dataset with diverse data types, including integers, strings, characters, booleans, and random criticality and priority bits. Experiments conducted in MATLAB demonstrate that CONTEXT-NET achieves stable throughput with a stability percentage of 94.72 %, while improving delivery probability by 6.45 %,reduces one-hop transmission delay by 28 %, end-to-end delay dropping by 7.9 % and mean overhead decreases by 5.96 % as the network scales from 50 to 100 nodes. These results confirm CONTEXT-NET’s ability to maintain consistent performance, enhance reliability, and improve efficiency in large-scale opportunistic networks. Validated across multiple application domains using a customized dataset with diverse data types and criticality levels, CONTEXT-NET emerges as a robust solution for real-world IoT and opportunistic networking applications.

机会网络使用存储-前转的方法在动态的、资源受限的环境中实现通信。然而，诸如有效的数据聚合、避免冲突、最小化数据冗余和信任管理等挑战仍然存在。本研究提出了上下文感知nexus聚合协议（CONTEXT-NET），该协议集成了空间、时间和上下文维度，以优化数据传输。CONTEXT-NET采用了一种具有同步扇区调度、基于自编码器的降维和一种类似蚁群优化（ACO）的混合路由算法的连接环拓扑结构，用于自适应路由，确保最小的冲突和有效的数据聚合。基于信任的评分系统通过识别和排除不可靠的节点来增强安全性。用于分析的数据集由自定义的随机数据集组成，该数据集具有多种数据类型，包括整数、字符串、字符、布尔值以及随机的临界位和优先位。在MATLAB中进行的实验表明，当网络从50个节点扩展到100个节点时，CONTEXT-NET实现了稳定的吞吐量，稳定率为94.72%，同时交付概率提高了6.45%，一跳传输延迟减少了28%，端到端延迟下降了7.9%，平均开销减少了5.96%。这些结果证实了CONTEXT-NET在大规模机会网络中保持一致性能、增强可靠性和提高效率的能力。使用具有不同数据类型和临界级别的自定义数据集在多个应用领域进行验证，CONTEXT-NET成为现实世界物联网和机会性网络应用的强大解决方案。

{"title":"CONTEXT-NET: A context-aware nexus-based aggregation protocol for opportunistic networks","authors":"Rounak Raman , Ayush Yadav , Deepika Kukreja , Deepak Kumar Sharma","doi":"10.1016/j.iot.2025.101809","DOIUrl":"10.1016/j.iot.2025.101809","url":null,"abstract":"<div><div>Opportunistic Networks enable communication in dynamic, resource-constrained environments using a store-carry-forward approach. However, challenges such as efficient data aggregation, collision avoidance, minimizing data redundancy, and trust management persist. This study proposes the Context-Aware Nexus-Based Aggregation Protocol (CONTEXT-NET), which integrates spatial, temporal, and contextual dimensions for optimized data transmission. CONTEXT-NET employs a nexus ring topology with synchronized sector-based scheduling, autoencoder-based dimensionality reduction, and a hybridized Ant Colony Optimization (ACO)-like routing algorithm for adaptive routing, ensuring minimal collisions and efficient data aggregation. A trust-based scoring system enhances security by identifying and excluding unreliable nodes. The dataset for analysis consists of a customized random dataset with diverse data types, including integers, strings, characters, booleans, and random criticality and priority bits. Experiments conducted in MATLAB demonstrate that CONTEXT-NET achieves stable throughput with a stability percentage of 94.72 %, while improving delivery probability by 6.45 %,reduces one-hop transmission delay by 28 %, end-to-end delay dropping by 7.9 % and mean overhead decreases by 5.96 % as the network scales from 50 to 100 nodes. These results confirm CONTEXT-NET’s ability to maintain consistent performance, enhance reliability, and improve efficiency in large-scale opportunistic networks. Validated across multiple application domains using a customized dataset with diverse data types and criticality levels, CONTEXT-NET emerges as a robust solution for real-world IoT and opportunistic networking applications.</div></div>","PeriodicalId":29968,"journal":{"name":"Internet of Things","volume":"34 ","pages":"Article 101809"},"PeriodicalIF":7.6,"publicationDate":"2025-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145362588","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}

引用次数: 0

eXplicability AI (XAI) for attack detection toward smart rural applications 可解释性AI （XAI）用于智能农村应用的攻击检测

IF 7.6 3区计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Internet of Things

Pub Date : 2025-10-20 DOI: 10.1016/j.iot.2025.101804

Enrique Fernández-Morales , Llanos Tobarra , Antonio Robles-Gómez , Rafael Pastor-Vargas , Roberto Hernández , Joao Sarraipa

This research evaluates the performance and computational efficiency of various AI models for intrusion detection in IoT environments, with the goal of enabling future deployment in Smart Rural scenarios. Leveraging the massive NF-UQ-NIDS-v2 dataset-comprising over 76 million labeled NetFlow records across 21 traffic classes-we benchmark five models, ranging from classical machine learning algorithms to deep learning architectures, across both high-performance and low-performance execution setups. The analysis covers standard classification metrics (accuracy, precision, recall, F1-score) and detailed resource usage indicators, including inference time, memory footprint, CPU cycles, and energy consumption per batch. Additionally, explainable AI techniques (SHAP and LIME) are employed to investigate model behavior and feature relevance under real-world constraints. Results show that classical models, particularly Random Forest and Decision Tree, achieve top-tier detection accuracy while maintaining minimal computational demands, making them strong candidates for constrained deployments. Deep learning models deliver comparable predictive performance but incur significantly higher resource consumption, requiring further optimization for practical use. Overall, this work provides a comprehensive evaluation framework and practical insights for selecting efficient and interpretable AI-based intrusion detection systems for rural and low-resource infrastructures.

本研究评估了物联网环境中用于入侵检测的各种人工智能模型的性能和计算效率，旨在实现未来在智慧农村场景中的部署。利用大规模的NF-UQ-NIDS-v2数据集，包括超过7600万条标记NetFlow记录，跨越21个流量类别，我们对五种模型进行基准测试，从经典机器学习算法到深度学习架构，跨越高性能和低性能执行设置。该分析涵盖标准分类指标（准确性、精密度、召回率、f1分数）和详细的资源使用指标，包括推理时间、内存占用、CPU周期和每批能耗。此外，可解释的人工智能技术（SHAP和LIME）被用于研究现实世界约束下的模型行为和特征相关性。结果表明，经典模型，特别是随机森林和决策树，在保持最小计算需求的同时实现了顶级检测精度，使其成为约束部署的有力候选者。深度学习模型提供了相当的预测性能，但会导致更高的资源消耗，需要进一步优化实际使用。总的来说，这项工作为为农村和低资源基础设施选择高效和可解释的基于人工智能的入侵检测系统提供了一个全面的评估框架和实践见解。

{"title":"eXplicability AI (XAI) for attack detection toward smart rural applications","authors":"Enrique Fernández-Morales , Llanos Tobarra , Antonio Robles-Gómez , Rafael Pastor-Vargas , Roberto Hernández , Joao Sarraipa","doi":"10.1016/j.iot.2025.101804","DOIUrl":"10.1016/j.iot.2025.101804","url":null,"abstract":"<div><div>This research evaluates the performance and computational efficiency of various AI models for intrusion detection in IoT environments, with the goal of enabling future deployment in Smart Rural scenarios. Leveraging the massive NF-UQ-NIDS-v2 dataset-comprising over 76 million labeled NetFlow records across 21 traffic classes-we benchmark five models, ranging from classical machine learning algorithms to deep learning architectures, across both high-performance and low-performance execution setups. The analysis covers standard classification metrics (accuracy, precision, recall, F1-score) and detailed resource usage indicators, including inference time, memory footprint, CPU cycles, and energy consumption per batch. Additionally, explainable AI techniques (SHAP and LIME) are employed to investigate model behavior and feature relevance under real-world constraints. Results show that classical models, particularly Random Forest and Decision Tree, achieve top-tier detection accuracy while maintaining minimal computational demands, making them strong candidates for constrained deployments. Deep learning models deliver comparable predictive performance but incur significantly higher resource consumption, requiring further optimization for practical use. Overall, this work provides a comprehensive evaluation framework and practical insights for selecting efficient and interpretable AI-based intrusion detection systems for rural and low-resource infrastructures.</div></div>","PeriodicalId":29968,"journal":{"name":"Internet of Things","volume":"34 ","pages":"Article 101804"},"PeriodicalIF":7.6,"publicationDate":"2025-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145362585","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}

引用次数: 0

Optimized K-means routing protocol with black-winged kite algorithm for sustainable 5G/6G sensor networks 基于黑翼风筝算法的可持续5G/6G传感器网络K-means路由协议优化

IF 7.6 3区计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Internet of Things

Pub Date : 2025-10-17 DOI: 10.1016/j.iot.2025.101792

Nishu Gupta , Mauro Mazzei , Jukka Mäkelä , Mikko Uitto

Wireless sensor networks (WSNs) face critical challenges due to energy-constrained nodes, affecting their longevity, reliability, and efficiency. To improve energy effectiveness in WSNs for fifth-generation and sixth-generation (5G/6G) networks, various clustering techniques have been developed. These techniques aim to optimize energy use, ensuring better system performance. Moreover, to overcome these complications, this article proposes a K-means online-learning routing protocol optimized with the black-winged kite optimization algorithm for sustainable communication (KORP-BWKOA-SC-WSN). Initially, the input data is collected from the sink node. This data is fed to a binarized simplicial convolutional neural network for cluster formation, in which the network nodes are clustered. Next, the formed cluster is used for cluster head selection by using the hiking optimization algorithm for better data transmission. Finally, the K-means online learning routing protocol is implemented to improve node coordination and energy efficiency. The black-winged kite optimization approach is employed to enhance the system performance. The proposed KORP-BWKOA-SC-WSN achieves throughput improvements of 21.51%, 12.38%, and 21.51%, respectively, and energy consumption reductions of 15.85%, 23.37%, and 22.04% compared to existing methods The performance of the proposed technique is evaluated and is found to attain higher throughput and high network lifetime when compared with other existing methods.

无线传感器网络（wsn）由于节点能量受限而面临严峻挑战，影响了其寿命、可靠性和效率。为了提高第五代和第六代（5G/6G）网络的无线传感器网络的能源效率，开发了各种聚类技术。这些技术旨在优化能源使用，确保更好的系统性能。此外，为了克服这些复杂性，本文提出了一种基于可持续通信黑翼风筝优化算法（KORP-BWKOA-SC-WSN）的k均值在线学习路由协议。最初，从汇聚节点收集输入数据。这些数据被馈送到二值化的简单卷积神经网络中进行聚类，其中网络节点被聚类。其次，将形成的聚类利用徒步优化算法进行簇头选择，以获得更好的数据传输。最后，实现K-means在线学习路由协议，提高节点协调和能量效率。采用黑翼风筝优化方法提高系统性能。与现有方法相比，所提出的KORP-BWKOA-SC-WSN的吞吐量分别提高了21.51%、12.38%和21.51%，能耗分别降低了15.85%、23.37%和22.04%。对所提出技术的性能进行了评估，发现与其他现有方法相比，该技术具有更高的吞吐量和更高的网络寿命。

{"title":"Optimized K-means routing protocol with black-winged kite algorithm for sustainable 5G/6G sensor networks","authors":"Nishu Gupta , Mauro Mazzei , Jukka Mäkelä , Mikko Uitto","doi":"10.1016/j.iot.2025.101792","DOIUrl":"10.1016/j.iot.2025.101792","url":null,"abstract":"<div><div>Wireless sensor networks (WSNs) face critical challenges due to energy-constrained nodes, affecting their longevity, reliability, and efficiency. To improve energy effectiveness in WSNs for fifth-generation and sixth-generation (5G/6G) networks, various clustering techniques have been developed. These techniques aim to optimize energy use, ensuring better system performance. Moreover, to overcome these complications, this article proposes a K-means online-learning routing protocol optimized with the black-winged kite optimization algorithm for sustainable communication (KORP-BWKOA-SC-WSN). Initially, the input data is collected from the sink node. This data is fed to a binarized simplicial convolutional neural network for cluster formation, in which the network nodes are clustered. Next, the formed cluster is used for cluster head selection by using the hiking optimization algorithm for better data transmission. Finally, the K-means online learning routing protocol is implemented to improve node coordination and energy efficiency. The black-winged kite optimization approach is employed to enhance the system performance. The proposed KORP-BWKOA-SC-WSN achieves throughput improvements of 21.51%, 12.38%, and 21.51%, respectively, and energy consumption reductions of 15.85%, 23.37%, and 22.04% compared to existing methods The performance of the proposed technique is evaluated and is found to attain higher throughput and high network lifetime when compared with other existing methods.</div></div>","PeriodicalId":29968,"journal":{"name":"Internet of Things","volume":"34 ","pages":"Article 101792"},"PeriodicalIF":7.6,"publicationDate":"2025-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145324442","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}

引用次数: 0

Efficient irrigation system using a combined wireless sensor network based on LoRaWAN and IEEE 802.15.4 technologies and photosynthetically active radiation measurements 利用基于LoRaWAN和IEEE 802.15.4技术的组合无线传感器网络和光合有效辐射测量的高效灌溉系统

IF 7.6 3区计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Internet of Things

Pub Date : 2025-10-17 DOI: 10.1016/j.iot.2025.101801

J. Medina-García , J.A. Gómez-Galán , J.M. Vilaplana-Guerrero , J.A. Bogeat

To address the demands of wireless communication, small amount of transmission, low power consumption, and cost-effectiveness in agricultural Internet of Things (IoT) applications, this paper introduces a hybrid information monitoring approach. It combines a low-data-rate personal area network based on IEEE 802.15.4 with a low-power wide-area network utilizing LoRaWAN. This method employs a communication architecture comprising a central node, multiple subnodes, and end devices to support the needs of large-scale information monitoring. Specifically, the main node is designed using LoRaWAN communication technology and performs in-field measurements of photosynthetically active radiation (PAR) using a device calibrated through intercomparison with reference radiometers. The subnodes or cluster heads incorporate LoRaWAN, sensor technologies, and IEEE 802.15.4. End devices also utilize IEEE 802.15.4 and sensor technologies. A control terminal manages sensor data and transmits the collected information to a web application for further processing. The advantages of this approach are that combining IEEE 802.15.4 and LoRaWAN at the device level enhances the spatial variability of agricultural fields, since tree and star network topologies are integrated to collect detailed information about specific crop areas, while providing low-power, long-distance network services and reducing the operating costs of the wide-area network information monitoring system. Additionally, hardware and firmware strategies were applied to further extend the system autonomy, and it can be self-powered. System testing revealed that, in a challenging environment, the maximum communication range reaches up to 60 m for IEEE 802.15.4 and 2 km for LoRaWAN. The average energy consumption is only 0.55 mAh, supporting real-time monitoring with latency under 100 ms, and the packet loss rate is approximately 2.5 %. Overall, the system operates reliably, and the data collected are accurate. The findings indicate that the proposed method effectively fulfills the needs for data gathering, transmission, storage, and processing across large areas. Furthermore, it proves to be valuable for implementing strategies aimed at improving both irrigation systems and the cultivation process of strawberry crops.

针对农业物联网应用中无线通信、传输量小、功耗低、性价比高等要求，提出了一种混合信息监控方法。它结合了基于IEEE 802.15.4的低数据速率个人局域网和利用LoRaWAN的低功耗广域网。该方法采用由一个中心节点、多个子节点和终端设备组成的通信体系结构，以支持大规模信息监控的需求。具体来说，主节点采用LoRaWAN通信技术设计，并使用通过与参考辐射计相互比较校准的设备进行光合有效辐射（PAR）的现场测量。子节点或簇头结合了LoRaWAN、传感器技术和IEEE 802.15.4。终端设备还使用IEEE 802.15.4和传感器技术。控制终端对传感器数据进行管理，并将采集到的信息发送给web应用程序进行进一步处理。该方法的优点是在设备层面将IEEE 802.15.4和LoRaWAN相结合，增强了农田的空间变异性，因为树形和星形网络拓扑相结合，可以收集特定作物区域的详细信息，同时提供低功耗、长距离的网络服务，降低了广域网信息监控系统的运行成本。此外，硬件和固件策略被应用于进一步扩展系统的自主性，它可以自供电。系统测试表明，在具有挑战性的环境下，IEEE 802.15.4的最大通信距离可达60米，LoRaWAN的最大通信距离可达2公里。平均功耗仅为0.55 mAh，支持100ms以下的实时监控，丢包率约为2.5%。总体而言，系统运行可靠，采集数据准确。研究结果表明，该方法有效地满足了大范围数据采集、传输、存储和处理的需求。此外，它被证明对实施旨在改善灌溉系统和草莓作物栽培过程的战略是有价值的。

{"title":"Efficient irrigation system using a combined wireless sensor network based on LoRaWAN and IEEE 802.15.4 technologies and photosynthetically active radiation measurements","authors":"J. Medina-García , J.A. Gómez-Galán , J.M. Vilaplana-Guerrero , J.A. Bogeat","doi":"10.1016/j.iot.2025.101801","DOIUrl":"10.1016/j.iot.2025.101801","url":null,"abstract":"<div><div>To address the demands of wireless communication, small amount of transmission, low power consumption, and cost-effectiveness in agricultural Internet of Things (IoT) applications, this paper introduces a hybrid information monitoring approach. It combines a low-data-rate personal area network based on IEEE 802.15.4 with a low-power wide-area network utilizing LoRaWAN. This method employs a communication architecture comprising a central node, multiple subnodes, and end devices to support the needs of large-scale information monitoring. Specifically, the main node is designed using LoRaWAN communication technology and performs in-field measurements of photosynthetically active radiation (PAR) using a device calibrated through intercomparison with reference radiometers. The subnodes or cluster heads incorporate LoRaWAN, sensor technologies, and IEEE 802.15.4. End devices also utilize IEEE 802.15.4 and sensor technologies. A control terminal manages sensor data and transmits the collected information to a web application for further processing. The advantages of this approach are that combining IEEE 802.15.4 and LoRaWAN at the device level enhances the spatial variability of agricultural fields, since tree and star network topologies are integrated to collect detailed information about specific crop areas, while providing low-power, long-distance network services and reducing the operating costs of the wide-area network information monitoring system. Additionally, hardware and firmware strategies were applied to further extend the system autonomy, and it can be self-powered. System testing revealed that, in a challenging environment, the maximum communication range reaches up to 60 m for IEEE 802.15.4 and 2 km for LoRaWAN. The average energy consumption is only 0.55 mAh, supporting real-time monitoring with latency under 100 ms, and the packet loss rate is approximately 2.5 %. Overall, the system operates reliably, and the data collected are accurate. The findings indicate that the proposed method effectively fulfills the needs for data gathering, transmission, storage, and processing across large areas. Furthermore, it proves to be valuable for implementing strategies aimed at improving both irrigation systems and the cultivation process of strawberry crops.</div></div>","PeriodicalId":29968,"journal":{"name":"Internet of Things","volume":"34 ","pages":"Article 101801"},"PeriodicalIF":7.6,"publicationDate":"2025-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145362586","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}

引用次数: 0

SWAST KHOJ: An IoT-driven real-time health monitoring system prototype 物联网驱动的实时健康监测系统原型

IF 7.6 3区计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Internet of Things

Pub Date : 2025-10-13 DOI: 10.1016/j.iot.2025.101796

Ramesh Saha , Sanjeev Kumar Bharadwaj , Sohail Saif , Suparna Biswas , Mohan Bansal , Prateek Jain , Linga Reddy Cenkeramaddi

This study focuses on developing a health monitoring system prototype using the Internet of Medical Things (IoMT) to provide continuous and real-time tracking of vital signs. The advent of IoT-enabled healthcare equipment is drastically changing the landscape of healthcare infrastructure, facilitating seamless communication among diverse devices and platforms. In this context, the significance of smart health monitoring applications in IoT has led to the exploration of various IoT frameworks by researchers. In this work, we have deployed a Wireless Body Area Network (WBAN) architecture in real-time healthcare prototype using discrete heterogeneous sensor nodes. The purpose of this prototype is to collect vital signs, which include temperature, heart rate, peripheral oxygen saturation (SpO2), and electrocardiogram (ECG). A Graphical User Interface (GUI) is used to display the gathered data in accordance with the Modified Early Warning Score (MEWS) medical criteria also risk level classification using k-Nearest Neighbor algorithm (kNN) based trained model. Periodically, our prototype, called SWAST KHOJ, uploads the data together with the timestamp to a local server. Additionally, we evaluate the prototype’s Quality of Service (QoS) parameter in an interior setting using a variety of short-range communication techniques, such as Bluetooth, ZigBee, and Local Area Network (LAN) or wired connection. Our evaluation reveals end-to-end delays of 0.514 ms for wired communication, 0.62 ms for Wi-Fi, 0.417 ms for ZigBee, and 1.92 ms for Bluetooth. Notably, our method demonstrates lower end-to-end delays compared to previous studies. In addition, we test the prototype’s throughput with several connection protocols.

本研究的重点是利用医疗物联网（IoMT）开发一种健康监测系统原型，以提供连续和实时的生命体征跟踪。支持物联网的医疗保健设备的出现正在彻底改变医疗保健基础设施的格局，促进了不同设备和平台之间的无缝通信。在此背景下，智能健康监测在物联网中的应用意义促使研究人员探索各种物联网框架。在这项工作中，我们在使用离散异构传感器节点的实时医疗原型中部署了无线体域网络（WBAN）架构。这个原型的目的是收集生命体征，包括温度、心率、外周氧饱和度（SpO2）和心电图（ECG）。使用图形用户界面（GUI）显示收集到的数据，这些数据是根据修正早期预警评分（MEWS）医学标准以及基于k-最近邻算法（kNN）的训练模型进行风险等级分类的。我们的原型（称为SWAST KHOJ）会定期将数据与时间戳一起上传到本地服务器。此外，我们使用各种短距离通信技术（如蓝牙、ZigBee和局域网（LAN）或有线连接）在内部设置中评估原型的服务质量（QoS）参数。我们的评估显示，有线通信的端到端延迟为0.514 ms， Wi-Fi为0.62 ms， ZigBee为0.417 ms，蓝牙为1.92 ms。值得注意的是，与之前的研究相比，我们的方法显示了更低的端到端延迟。此外，我们用几种连接协议测试了原型的吞吐量。

{"title":"SWAST KHOJ: An IoT-driven real-time health monitoring system prototype","authors":"Ramesh Saha , Sanjeev Kumar Bharadwaj , Sohail Saif , Suparna Biswas , Mohan Bansal , Prateek Jain , Linga Reddy Cenkeramaddi","doi":"10.1016/j.iot.2025.101796","DOIUrl":"10.1016/j.iot.2025.101796","url":null,"abstract":"<div><div>This study focuses on developing a health monitoring system prototype using the Internet of Medical Things (IoMT) to provide continuous and real-time tracking of vital signs. The advent of IoT-enabled healthcare equipment is drastically changing the landscape of healthcare infrastructure, facilitating seamless communication among diverse devices and platforms. In this context, the significance of smart health monitoring applications in IoT has led to the exploration of various IoT frameworks by researchers. In this work, we have deployed a Wireless Body Area Network (WBAN) architecture in real-time healthcare prototype using discrete heterogeneous sensor nodes. The purpose of this prototype is to collect vital signs, which include temperature, heart rate, peripheral oxygen saturation (SpO2), and electrocardiogram (ECG). A Graphical User Interface (GUI) is used to display the gathered data in accordance with the Modified Early Warning Score (MEWS) medical criteria also risk level classification using k-Nearest Neighbor algorithm (kNN) based trained model. Periodically, our prototype, called SWAST KHOJ, uploads the data together with the timestamp to a local server. Additionally, we evaluate the prototype’s Quality of Service (QoS) parameter in an interior setting using a variety of short-range communication techniques, such as Bluetooth, ZigBee, and Local Area Network (LAN) or wired connection. Our evaluation reveals end-to-end delays of 0.514 ms for wired communication, 0.62 ms for Wi-Fi, 0.417 ms for ZigBee, and 1.92 ms for Bluetooth. Notably, our method demonstrates lower end-to-end delays compared to previous studies. In addition, we test the prototype’s throughput with several connection protocols.</div></div>","PeriodicalId":29968,"journal":{"name":"Internet of Things","volume":"34 ","pages":"Article 101796"},"PeriodicalIF":7.6,"publicationDate":"2025-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145324440","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}

引用次数: 0

Contemporary smart hydroponics systems: Taxonomy, enabling technologies, and challenges 当代智能水培系统：分类、实现技术和挑战

IF 7.6 3区计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Internet of Things

Pub Date : 2025-10-11 DOI: 10.1016/j.iot.2025.101794

Abeer Ahmed , Rawan Alnawasrah , Basem Almadani , Farouq Aliyu , Mustafa Ghaleb

Hydroponic farming is a sustainable agricultural technique that enables plant cultivation without soil. Hydroponics has evolved into a variety of configurations that integrate smart technologies to improve its efficiency and productivity. This paper reviews 40 publications to investigate the contemporary technologies, advantages, and challenges associated with modern hydroponic and aquaponic systems. Recent studies emphasize the role of middleware, Artificial Intelligence (AI), the Internet of Things (IoT), and the Industrial Internet of Things (IIoT) in enhancing system functionality. Middleware plays a critical role in facilitating seamless communication between system components. It also enables real-time capabilities through various communication protocols with a focus on quality of service (QoS). IoT and IIoT technologies enable data collection and environmental control, while AI contributes to automation, predictive analytics, and decision support. Together, these technologies help reduce resource consumption, such as nutrients, water, and energy. It also enables scalable, adaptive, and sustainable farming practices. Some of the challenges associated with hydroponic systems include initial development costs, security and privacy concerns, and the complexity of integrating advanced middleware.

水培农业是一种可持续的农业技术，可以在没有土壤的情况下种植植物。水培法已经发展成各种各样的配置，集成了智能技术，以提高其效率和生产力。本文回顾了40篇出版物，探讨了与现代水培和水耕系统相关的当代技术、优势和挑战。最近的研究强调中间件、人工智能（AI）、物联网（IoT）和工业物联网（IIoT）在增强系统功能方面的作用。中间件在促进系统组件之间的无缝通信方面起着关键作用。它还通过关注服务质量（QoS）的各种通信协议实现实时功能。物联网和工业物联网技术实现了数据收集和环境控制，而人工智能有助于自动化、预测分析和决策支持。总之，这些技术有助于减少资源消耗，如养分、水和能源。它还使可扩展、适应性强和可持续的农业实践成为可能。与水培系统相关的一些挑战包括初始开发成本、安全性和隐私问题，以及集成高级中间件的复杂性。

{"title":"Contemporary smart hydroponics systems: Taxonomy, enabling technologies, and challenges","authors":"Abeer Ahmed , Rawan Alnawasrah , Basem Almadani , Farouq Aliyu , Mustafa Ghaleb","doi":"10.1016/j.iot.2025.101794","DOIUrl":"10.1016/j.iot.2025.101794","url":null,"abstract":"<div><div>Hydroponic farming is a sustainable agricultural technique that enables plant cultivation without soil. Hydroponics has evolved into a variety of configurations that integrate smart technologies to improve its efficiency and productivity. This paper reviews 40 publications to investigate the contemporary technologies, advantages, and challenges associated with modern hydroponic and aquaponic systems. Recent studies emphasize the role of middleware, Artificial Intelligence (AI), the Internet of Things (IoT), and the Industrial Internet of Things (IIoT) in enhancing system functionality. Middleware plays a critical role in facilitating seamless communication between system components. It also enables real-time capabilities through various communication protocols with a focus on quality of service (QoS). IoT and IIoT technologies enable data collection and environmental control, while AI contributes to automation, predictive analytics, and decision support. Together, these technologies help reduce resource consumption, such as nutrients, water, and energy. It also enables scalable, adaptive, and sustainable farming practices. Some of the challenges associated with hydroponic systems include initial development costs, security and privacy concerns, and the complexity of integrating advanced middleware.</div></div>","PeriodicalId":29968,"journal":{"name":"Internet of Things","volume":"34 ","pages":"Article 101794"},"PeriodicalIF":7.6,"publicationDate":"2025-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145324441","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}

引用次数: 0

Federated learning with dual dynamic quantization optimization in smart agriculture 基于双动态量化优化的智能农业联邦学习

IF 7.6 3区计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Internet of Things

Pub Date : 2025-10-11 DOI: 10.1016/j.iot.2025.101798

Shaohui Zhang , Qiuying Han , Hongfeng Wang , Jing Liu , Boyuan Li

The convergence of Federated Learning (FL), Machine Learning (ML), and the Internet of Things (IoT) creates promising opportunities for smart agriculture, where connectivity constraints and limited device resources pose major bottlenecks. To address these challenges, we propose a Dual Dynamic Quantization Optimization (FedDDO) framework that jointly integrates quantizer design, adaptive bit allocation, and quantization-error-aware aggregation. On the client side, FedDDO dynamically adjusts quantization bit-widths according to real-time resource conditions, while on the server side, aggregation weights are optimized based on quantization error feedback. A novel Minimum Relative Quantization Error (MRQE) quantizer is designed to align with unbiased error assumptions, and theoretical analysis under non-convex settings provides convergence guarantees. Extensive experiments on both standard benchmarks (CIFAR-10/100) and agriculture-specific datasets (rice seedling classification and disease recognition) demonstrate that FedDDO effectively reduces communication costs and accelerates convergence, achieving competitive accuracy while preserving domain applicability.

联邦学习（FL）、机器学习（ML）和物联网（IoT）的融合为智能农业创造了充满希望的机会，在智能农业中，连接限制和有限的设备资源构成了主要瓶颈。为了解决这些挑战，我们提出了一个双动态量化优化（FedDDO）框架，该框架联合集成了量化器设计、自适应比特分配和量化错误感知聚合。在客户端，FedDDO根据实时资源情况动态调整量化位宽，在服务器端，基于量化误差反馈优化聚合权值。设计了一种新的最小相对量化误差（MRQE）量化器，以对准无偏误差假设，并在非凸设置下的理论分析提供了收敛保证。在标准基准（CIFAR-10/100）和农业特定数据集（水稻幼苗分类和疾病识别）上进行的大量实验表明，FedDDO有效地降低了通信成本并加速了收敛，在保持领域适用性的同时获得了具有竞争力的准确性。

{"title":"Federated learning with dual dynamic quantization optimization in smart agriculture","authors":"Shaohui Zhang , Qiuying Han , Hongfeng Wang , Jing Liu , Boyuan Li","doi":"10.1016/j.iot.2025.101798","DOIUrl":"10.1016/j.iot.2025.101798","url":null,"abstract":"<div><div>The convergence of Federated Learning (FL), Machine Learning (ML), and the Internet of Things (IoT) creates promising opportunities for smart agriculture, where connectivity constraints and limited device resources pose major bottlenecks. To address these challenges, we propose a Dual Dynamic Quantization Optimization (FedDDO) framework that jointly integrates quantizer design, adaptive bit allocation, and quantization-error-aware aggregation. On the client side, FedDDO dynamically adjusts quantization bit-widths according to real-time resource conditions, while on the server side, aggregation weights are optimized based on quantization error feedback. A novel Minimum Relative Quantization Error (MRQE) quantizer is designed to align with unbiased error assumptions, and theoretical analysis under non-convex settings provides convergence guarantees. Extensive experiments on both standard benchmarks (CIFAR-10/100) and agriculture-specific datasets (rice seedling classification and disease recognition) demonstrate that FedDDO effectively reduces communication costs and accelerates convergence, achieving competitive accuracy while preserving domain applicability.</div></div>","PeriodicalId":29968,"journal":{"name":"Internet of Things","volume":"34 ","pages":"Article 101798"},"PeriodicalIF":7.6,"publicationDate":"2025-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145324443","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}

引用次数: 0

IoT enabled health indicators estimation and indoor environment classification 基于物联网的健康指标评估和室内环境分类

IF 7.6 3区计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Internet of Things

Pub Date : 2025-10-10 DOI: 10.1016/j.iot.2025.101791

Cezar Anicai, Muhammad Zeeshan Shakir

Internet of Things (IoT) and Machine Learning (ML) have revolutionized the way we approach monitoring and analysing physiological data. Through these technologies invaluable insights can be gathered for early detection of cardiovascular issues, optimizing exercise routines or predicting stress levels. This study presents the development of an IoT test-bed, utilizing a single-board computer alongside ambient environment and health sensors for data collection. A data analysis pipeline has been designed to accurately estimate Heart Rate (HR) and Skin Resistance (SR) values exclusively using the ambient environment data and to classify the environment according to the risk it poses on cardiac health. The results of this study indicate the potential of using ML to capture the relationships between ambient environment conditions and health indicators. It has been found that Random Forest (RF) models are capable of classifying environments in three risk categories with an accuracy of 86.5% and estimate HR and SR with a MAE of 1.86 and 0.36, respectively. These contributions collectively advance the understanding of how environmental factors such as temperature, humidity, pressure and air quality influence health and show a promising potential for non-invasive monitoring.

物联网（IoT）和机器学习（ML）彻底改变了我们监测和分析生理数据的方式。通过这些技术，可以收集宝贵的见解，以早期发现心血管问题，优化锻炼程序或预测压力水平。本研究介绍了物联网试验台的开发，利用单板计算机以及环境环境和健康传感器进行数据收集。设计了一个数据分析管道，专门使用环境数据准确估计心率（HR）和皮肤阻力（SR）值，并根据环境对心脏健康构成的风险对环境进行分类。这项研究的结果表明，使用机器学习来捕捉环境条件和健康指标之间的关系具有潜力。随机森林（Random Forest， RF）模型能够对三种风险类别的环境进行分类，准确率为86.5%，估计HR和SR的MAE分别为1.86和0.36。这些贡献共同促进了对温度、湿度、压力和空气质量等环境因素如何影响健康的理解，并显示出非侵入性监测的良好潜力。

{"title":"IoT enabled health indicators estimation and indoor environment classification","authors":"Cezar Anicai, Muhammad Zeeshan Shakir","doi":"10.1016/j.iot.2025.101791","DOIUrl":"10.1016/j.iot.2025.101791","url":null,"abstract":"<div><div>Internet of Things (IoT) and Machine Learning (ML) have revolutionized the way we approach monitoring and analysing physiological data. Through these technologies invaluable insights can be gathered for early detection of cardiovascular issues, optimizing exercise routines or predicting stress levels. This study presents the development of an IoT test-bed, utilizing a single-board computer alongside ambient environment and health sensors for data collection. A data analysis pipeline has been designed to accurately estimate Heart Rate (HR) and Skin Resistance (SR) values exclusively using the ambient environment data and to classify the environment according to the risk it poses on cardiac health. The results of this study indicate the potential of using ML to capture the relationships between ambient environment conditions and health indicators. It has been found that Random Forest (RF) models are capable of classifying environments in three risk categories with an accuracy of 86.5% and estimate HR and SR with a MAE of 1.86 and 0.36, respectively. These contributions collectively advance the understanding of how environmental factors such as temperature, humidity, pressure and air quality influence health and show a promising potential for non-invasive monitoring.</div></div>","PeriodicalId":29968,"journal":{"name":"Internet of Things","volume":"34 ","pages":"Article 101791"},"PeriodicalIF":7.6,"publicationDate":"2025-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145324445","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}

引用次数: 0

Measuring the impact of post quantum cryptography in Industrial IoT scenarios 测量后量子加密在工业物联网场景中的影响

IF 7.6 3区计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Internet of Things

Pub Date : 2025-10-10 DOI: 10.1016/j.iot.2025.101793

Luis Cruz-Piris , Andrés Marín-López , Manuel Álvarez-Campana , Mario Sanz , José Ignacio Moreno , David Arroyo

The continuously evolving nature of cryptography is driven by the emergence of new threats and attack vectors. Quantum computers pose a paradigmatic security risk to cryptography, challenging its very core principles. This quantum threat can be appropriately addressed through quantum-safe cryptographic primitives, such as quantum key distribution and post-quantum cryptography (PQC). In the case of PQC, the paradigm shift involves using algorithms with significantly higher computational costs. This paper analyzes the possibilities and challenges of transitioning from current cryptographic systems to PQC alternatives, with a focus on the critical case of constrained-resource devices. We demonstrate the feasibility of such a transition in IoT and Industrial IoT (IIoT) scenarios with limited nodes, and we evaluate how new proposals can mitigate the impact of signature computations on securing IoT/IIoT devices. In this work, we design and implement a novel framework to conduct an extensive set of experiments measuring the performance of different families of PQC algorithms in terms of execution time and power consumption. Both the framework and the dataset have been published in the EU Open Research Repository Zenodo to facilitate the future selection of algorithms that best adapt to the specific characteristics of each system.

密码学不断发展的本质是由新的威胁和攻击媒介的出现所驱动的。量子计算机对密码学构成了典型的安全风险，挑战了密码学的核心原则。这种量子威胁可以通过量子安全密码原语（如量子密钥分发和后量子密码（PQC））适当地解决。在PQC的情况下，范式转换涉及使用具有显着更高计算成本的算法。本文分析了从当前密码系统过渡到PQC替代方案的可能性和挑战，重点讨论了受限资源设备的关键情况。我们论证了这种过渡在节点有限的物联网和工业物联网（IIoT）场景中的可行性，并评估了新提案如何减轻签名计算对保护物联网/工业物联网设备的影响。在这项工作中，我们设计并实现了一个新的框架，以进行一组广泛的实验，测量不同家族的PQC算法在执行时间和功耗方面的性能。框架和数据集都已在欧盟开放研究存储库Zenodo上发布，以方便将来选择最适合每个系统特定特征的算法。

{"title":"Measuring the impact of post quantum cryptography in Industrial IoT scenarios","authors":"Luis Cruz-Piris , Andrés Marín-López , Manuel Álvarez-Campana , Mario Sanz , José Ignacio Moreno , David Arroyo","doi":"10.1016/j.iot.2025.101793","DOIUrl":"10.1016/j.iot.2025.101793","url":null,"abstract":"<div><div>The continuously evolving nature of cryptography is driven by the emergence of new threats and attack vectors. Quantum computers pose a paradigmatic security risk to cryptography, challenging its very core principles. This quantum threat can be appropriately addressed through quantum-safe cryptographic primitives, such as quantum key distribution and post-quantum cryptography (PQC). In the case of PQC, the paradigm shift involves using algorithms with significantly higher computational costs. This paper analyzes the possibilities and challenges of transitioning from current cryptographic systems to PQC alternatives, with a focus on the critical case of constrained-resource devices. We demonstrate the feasibility of such a transition in IoT and Industrial IoT (IIoT) scenarios with limited nodes, and we evaluate how new proposals can mitigate the impact of signature computations on securing IoT/IIoT devices. In this work, we design and implement a novel framework to conduct an extensive set of experiments measuring the performance of different families of PQC algorithms in terms of execution time and power consumption. Both the framework and the dataset have been published in the EU Open Research Repository Zenodo to facilitate the future selection of algorithms that best adapt to the specific characteristics of each system.</div></div>","PeriodicalId":29968,"journal":{"name":"Internet of Things","volume":"34 ","pages":"Article 101793"},"PeriodicalIF":7.6,"publicationDate":"2025-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145324439","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}

引用次数: 0

Exploiting edge features for transferable adversarial attacks in distributed machine learning 利用边缘特征在分布式机器学习中进行可转移的对抗性攻击

IF 7.6 3区计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS

Internet of Things

Pub Date : 2025-10-10 DOI: 10.1016/j.iot.2025.101795

Giulio Rossolini , Fabio Brau , Alessandro Biondi , Battista Biggio , Giorgio Buttazzo

As machine learning models become increasingly deployed across the edge of internet of things environments, a partitioned deep learning paradigm in which models are split across multiple computational nodes introduces a new dimension of security risk. Unlike traditional inference setups, these distributed pipelines span the model computation across heterogeneous nodes and communication layers, thereby exposing a broader attack surface to potential adversaries. Building on these motivations, this work explores a previously overlooked vulnerability: even when both the edge and cloud components of the model are inaccessible (i.e., black-box), an adversary who intercepts the intermediate features transmitted between them can still pose a serious threat. We demonstrate that, under these mild and realistic assumptions, an attacker can craft highly transferable proxy models, making the entire deep learning system significantly more vulnerable to evasion attacks. In particular, the intercepted features can be effectively analyzed and leveraged to distill surrogate models capable of crafting highly transferable adversarial examples against the target model. To this end, we propose an exploitation strategy specifically designed for distributed settings, which involves reconstructing the original tensor shape from vectorized transmitted features using simple statistical analysis, and adapting surrogate architectures accordingly to enable effective feature distillation.

A comprehensive and systematic experimental evaluation has been conducted to demonstrate that surrogate models trained with the proposed strategy, i.e., leveraging intermediate features, tremendously improve the transferability of adversarial attacks. These findings underscore the urgent need to account for intermediate feature leakage in the design of secure distributed deep learning systems, particularly in edge scenarios, where constrained devices are more exposed to communication vulnerabilities and offer limited protection mechanisms.

随着机器学习模型越来越多地部署在物联网环境的边缘，一种分割的深度学习范式（模型在多个计算节点上分割）引入了一个新的安全风险维度。与传统的推理设置不同，这些分布式管道跨越异构节点和通信层的模型计算，从而向潜在的对手暴露更广泛的攻击面。在这些动机的基础上，这项工作探索了一个以前被忽视的漏洞：即使模型的边缘和云组件都是不可访问的（即黑盒），拦截它们之间传输的中间特征的对手仍然可以构成严重的威胁。我们证明，在这些温和而现实的假设下，攻击者可以制作高度可转移的代理模型，使整个深度学习系统更容易受到逃避攻击。特别是，可以有效地分析和利用截获的特征来提取代理模型，这些模型能够针对目标模型制作高度可转移的对抗性示例。为此，我们提出了一种专门为分布式设置设计的开发策略，该策略包括使用简单的统计分析从矢量化的传输特征中重建原始张量形状，并相应地调整代理架构以实现有效的特征蒸馏。已经进行了全面和系统的实验评估，以证明用所提出的策略训练的代理模型，即利用中间特征，极大地提高了对抗性攻击的可转移性。这些发现强调了迫切需要考虑安全分布式深度学习系统设计中的中间特征泄漏，特别是在边缘场景中，受约束的设备更容易暴露于通信漏洞并且提供有限的保护机制。

{"title":"Exploiting edge features for transferable adversarial attacks in distributed machine learning","authors":"Giulio Rossolini , Fabio Brau , Alessandro Biondi , Battista Biggio , Giorgio Buttazzo","doi":"10.1016/j.iot.2025.101795","DOIUrl":"10.1016/j.iot.2025.101795","url":null,"abstract":"<div><div>As machine learning models become increasingly deployed across the edge of internet of things environments, a partitioned deep learning paradigm in which models are split across multiple computational nodes introduces a new dimension of security risk. Unlike traditional inference setups, these distributed pipelines span the model computation across heterogeneous nodes and communication layers, thereby exposing a broader attack surface to potential adversaries. Building on these motivations, this work explores a previously overlooked vulnerability: even when both the edge and cloud components of the model are inaccessible (i.e., black-box), an adversary who intercepts the intermediate features transmitted between them can still pose a serious threat. We demonstrate that, under these mild and realistic assumptions, an attacker can craft highly transferable proxy models, making the entire deep learning system significantly more vulnerable to evasion attacks. In particular, the intercepted features can be effectively analyzed and leveraged to distill surrogate models capable of crafting highly transferable adversarial examples against the target model. To this end, we propose an exploitation strategy specifically designed for distributed settings, which involves reconstructing the original tensor shape from vectorized transmitted features using simple statistical analysis, and adapting surrogate architectures accordingly to enable effective feature distillation.</div><div>A comprehensive and systematic experimental evaluation has been conducted to demonstrate that surrogate models trained with the proposed strategy, i.e., leveraging intermediate features, tremendously improve the transferability of adversarial attacks. These findings underscore the urgent need to account for intermediate feature leakage in the design of secure distributed deep learning systems, particularly in edge scenarios, where constrained devices are more exposed to communication vulnerabilities and offer limited protection mechanisms.</div></div>","PeriodicalId":29968,"journal":{"name":"Internet of Things","volume":"34 ","pages":"Article 101795"},"PeriodicalIF":7.6,"publicationDate":"2025-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145324437","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}

引用次数: 0