TNSM Searchable Index

Author(s)	Title	Year	Publication	Keywords
Wei-Che Chien, Gwanggil Jeon, Hsin-Hung Cho	Multi-Objective Optimization of 3D Cell Deployment in Sustainable B5G/6G Networks: Balancing Performance and Sustainability	2025	Early Access	Base stations 5G mobile communication Wireless communication	In recent years, the exponential increase in mobile and Internet of Things (IoT) data traffic has placed substantial demands on infrastructure for Internet Service Providers (ISPs). To meet these demands sustainably, it is critical to enhance energy efficiency, resource utilization, and cost-effectiveness while reducing the carbon footprint. Simply adding hardware is not a viable solution. This study introduces an innovative approach to 3D cellular deployment in sustainable B5G/6G networks, designed to optimize Quality of Service (QoS) for users and IoT devices. Although 5G/B5G utilizes millimeter waves for high data rate transmission, their limited coverage and susceptibility to interference from buildings pose unique deployment challenges. To address these, we formulate the 3D cellular deployment problem as a Multi-Objective Optimization (MOO) problem and propose an advanced deployment strategy using VA-NSGA-II, a metaheuristic-based algorithm. By factoring in building interference, Received Signal Strength Indicator (RSSI), coverage, deployment cost, and a balance between performance and sustainability, VA-NSGA-II provides an optimal deployment solution. Simulation results demonstrate that VA-NSGA-II achieves effective deployment performance across various building materials, highlighting its adaptability and effectiveness in different environmental scenarios.	10.1109/TNSM.2025.3545622
Giuseppe Ruggeri, Marica Amadeo, Claudia Campolo, Antonella Molinaro	Optimal Placement of the Virtualized Federated Learning Aggregation Function at the Edge	2025	Early Access	Training Computational modeling Servers	Federated Learning (FL) enables multiple devices (clients) training a shared machine learning (ML) model on local datasets and then sending the updated models to a central server, whose task is aggregating the locally-computed updates and sharing the learned global model again with the clients in an iterative process. The population of clients may change at each round, whereas the node executing the aggregation function is typically placed at an edge domain and remains static until the end of the overall FL training process. Indeed, the computing capabilities of the edge node hosting the aggregation function and the distance (latency) of such a node from the selected clients can highly affect the convergence rate of the FL training procedure. Moreover, the heterogeneous time-varying capabilities of edge nodes, coupled with the dynamic client population selected at each round, call for the optimal dynamic placement of the aggregation function across the available nodes in an edge domain. In this work, we formulate an optimization problem for the placement of the FL aggregation function, which aims to select at each round the edge node able to minimize the overall per-round training time, encompassing the aggregation time, the local training time at the clients and the time for exchanging the global model and the model updates. A time-efficient greedy heuristics is proposed, which is shown to well approximate the optimal solution and outperform the considered benchmark solutions.	10.1109/TNSM.2025.3551257
Takanori Hara, Masahiro Sasabe	eBPF-Based Ordered Proof of Transit for Trustworthy Service Function Chaining	2025	Early Access	Security Routing Kernel	Service function chaining (SFC) establishes a service path where a sequence of functions is executed according to service requirements. However, SFC lacks a mechanism to ensure proper traversal of relay nodes in the data plane. Misconfigurations and the presence of attackers can lead to forwarding anomalies and path deviation, potentially allowing packets to bypass security network functions in the service path. To mitigate potential security breaches, ordered proof of transit (OPoT) has been proposed as a mechanism to verify whether traffic adheres to the designated path. In this paper, we realize lightweight OPoT-based path verification based on extended Berkeley Packet Filter (eBPF) for trustworthy SFC. Furthermore, by integrating it with the existing SFC proxy, we extend the proposed approach to accommodate both SFC-aware and SFC-unaware virtual network functions (VNFs) in the segment routing over IPv6 data plane (SRv6) domain. Through experiments, we demonstrate the capability of the proposed approach to detect path deviations. Additionally, we reveal the performance limitations of the proposed approach.	10.1109/TNSM.2025.3550333
Kai Peng, Jialu Guo, Hao Wang, Jintao He, Zhiqing Zou, Tianping Deng, Menglan Hu	Delay-Aware Joint Microservice Deployment and Request Routing in Multi-Edge Environments Based on Reinforcement Learning	2025	Early Access	Microservice architectures Routing Cloud computing	The service modules of the traditional Mobile Edge Computing (MEC) are difficult to deploy, extend, and maintain in real networks because of the highly sophisticated systems. To promote the generalization, openness, and flexibility of the network edge environment, an increasing number of studies are exploring the integration of microservices with MEC. However, the existing work usually treats microservice deployment and request routing as two separate issues, ignoring the interaction between them. Therefore, this paper focuses on the joint optimization of microservice deployment and request routing in the multi-edge cloud scenarios. We establish a problem model for minimizing the average response latency, considering the transmission of requests across edge clouds. Then, in view of the complexity of the scene, this paper proposes a joint training strategy of microservice deployment and request routing based on deep reinforcement learning and Best Fit Decreasing algorithm. The algorithm takes the change of microservice deployment scheme as the action of the agent, introduces the Best Fit Decreasing algorithm to construct request routing based on the deployment scheme, and calculates rewards using the complete joint microservice deployment and request routing scheme for subsequent network training. Finally, experimental results show that the proposed algorithm can effectively reduce the response time delay and system running power compared with other algorithms.	10.1109/TNSM.2025.3543568
Sławomir Hanczewski, Maciej Stasiak, Joanna Weissenberg, Michał Weissenberg	Modelling of Heterogeneous 5G Network Slice for Smart Real-Time Railway Communications	2025	Early Access	Real-time systems Rail transportation Analytical models	This paper presents an analytical model for a railway mobile communications system. In line with recent trends, the system’s operation relies on 5G network resources (slices). It efficiently manages critical data streams (flows) that meet the stringent requirements of real-time systems (systems that handle hard and soft real-time services). Additionally, the proposed solution accommodates data with less stringent QoS parameters compared to real-time streams. The analytical model serves as an approximation of the process occurring in the system for servicing flows and has been developed based on the analysis of a Markov chain, where the states correspond to the states of the examined system. Due to the approximate nature of the analytical model, the results derived from it were compared with those obtained from the simulation experiment.	10.1109/TNSM.2025.3547762
Mengjie Guo, Guochu Shou, Yaqiong Liu, Yihong Hu	An Enhanced Reconfiguration for Deterministic Transmission in Time-Sensitive Networks	2025	Early Access	Logic gates Heuristic algorithms Vehicle dynamics	Time-aware shaper (TAS) is key to enabling deterministic guarantees in time-sensitive networks (TSN), but it requires precise configuration for specific traffic scenarios. Dynamic traffic scenarios are increasingly commonplace with the rise of emerging applications, necessitating TAS reconfiguration to adapt to the changes in traffic. However, existing mechanisms primarily reconfigure TAS by generating a new gate control list (GCL) and transitioning to it, which may lead to temporary violations of bounds on delay or jitter, providing no persistently deterministic guarantees. In this paper, we propose a novel TAS reconfiguration mechanism with the virtual GCL (VGCL) to satisfy the demands of dynamic traffic while guaranteeing deterministic transmission. It implements TAS reconfiguration for dynamic traffic by embedding different VGCLs into the GCL, avoiding the need for the GCL transition. Thus, the reconfiguration problem is modeled as an embedding problem by using the VGCL and we develop algorithms to solve it. Experimental results demonstrate that our mechanism can well reconfigure TAS for dynamic traffic without the GCL transition, and increase the reconfiguration success rate in various scenarios compared with the existing approaches.	10.1109/TNSM.2025.3547896
Gerald Tietaa Maale, Noble Arden Elorm Kuadey, Yeasin Arafat, Kwantwi Thomas, Guolin Sun, Guisong Liu	Multi-Task Learning for UAV Trajectory and Caching with Federated Cloud-Assisted Knowledge Distillation	2025	Early Access	Autonomous aerial vehicles Optimization Trajectory	The proliferation of Internet of Things (IoT) technologies and ubiquitous connectivity has led to unmanned aerial vehicles (UAVs) playing key role as edge servers, revolutionizing the wireless communications landscape by facilitating computing and caching resources closer to ground users (GUs). This advancement significantly alleviates core network loads, reduces latency, and guarantees content availability even in congested or remote areas. However, jointly optimizing UAV caching strategies and trajectories gives rise to a multi-task optimization (MTO) problem. This paper introduces a novel multi-task geo-temporal caching (MT-GTC) framework that addresses the interplay between UAV caching mechanisms and trajectory optimization in a cohesive manner. Leveraging a proposed multi-task learning (MTL) model for joint optimization of UAV caching and trajectory design, we develop a federated learning cloud-assisted knowledge distillation (FL-CAKD) scheme to preserve data privacy and adapt to data heterogeneity. FL-CAKD transfers knowledge from a cloud model orchestrator (CMO), which houses a large and sophisticated teacher model, to a lightweight on-device MTL student models using soft target distributions instead of large model parameters, significantly reducing communication costs. MT-GTC optimizes caching and trajectories to maximize cache hits and minimize latency. Evaluations on real-world mobility datasets demonstrate up to 95% cache hit rates and 21% lower delays compared to baselines.	10.1109/TNSM.2025.3547743
Jing Mei, Cuibin Zeng, Zhao Tong, Zhibang Yang, Keqin Li	Stackelberg Game-Based Pricing and Offloading for the DVFS-Enabled MEC Systems	2025	Early Access	Servers Games Pricing	Due to the limited computing resources of both mobile devices (MDs) and the mobile edge computing (MEC) server, devising reasonable strategies for MD task offloading, MEC server resource pricing, and resource allocation is crucial. In this paper, a scenario is considered, comprising multiple MDs and a single MEC server. Each MD has a divisible task in each time slot, allowing for partial offloading and the option to discard parts of the task. The MEC server contains multiple computing units with the same computing power, and its computing resources can be dynamically adjusted through dynamic voltage and frequency scaling (DVFS) according to the size of tasks offloaded by MDs. At any given time slice, a Stackelberg game is formulated based on the strategies of the MDs and the strategy of the MEC server. An iterative evolution algorithm is employed to explore the optimal strategies for MDs and the MEC server. Simulation results demonstrate that both parties can reach an equilibrium state through the game, and these experiments confirm that the algorithm effectively enhances system efficiency.	10.1109/TNSM.2025.3547568
Jun Jiang, Bin Wang, Quan Tang, Guoxiang Zhong, Xuhao Tang, Joel J. P. C. Rodrigues	Incremental Semi-Supervised Learning for Data Streams Classification in Internet of Things	2025	Early Access	Streams Data models Heuristic algorithms	Data stream classification is widely used in Internet of Things (IoT) scenarios such as health monitoring, anomaly detection and online diagnosis. Due to the continuous data stream changing dynamically over time, it is impossible to classify all the data simultaneously. Moreover, labeling each sample in practical data stream applications is time-and resource-consuming. The realistic situation is that only a few instances in a data stream are labeled. Therefore, classifying data streams with limited labels has become challenging in IoT scenarios. In this paper, we propose an incremental dynamic weighted semi-supervised method for classifying IoT data streams. Considering the dynamics and continuity in data streams, we use a chunk-based approach to learn the features in the data stream and assign weights to the classifier dynamically. Moreover, we deploy incremental learning methods to continuously learn from the sampled labeled data stream to update the classifier model, which can take advantage of newly incoming labeled data to improve learning performance. Experimental evaluations on seven IoT datasets show that the proposed method outperforms semi-supervised methods in accuracy, precision, and geometric mean (Gmean) by 10% and 5% over supervised methods, respectively.	10.1109/TNSM.2025.3546649
Dev Gurung, Shiva Raj Pokhrel, Gang Li	Quantum Federated Learning for Metaverse: Analysis, Design and Implementation	2025	Early Access	Metaverse Blockchains Quantum computing	We present a novel decentralized and trustworthy Quantum Federated Learning (QFL) framework tailored for the emerging Metaverse. This virtual environment, enabling social interaction, gaming, and commerce, demands secure and transparent systems. By integrating blockchain, our QFL framework ensures integrity, resilience, and transparency. Comparative analysis with classical Federated Learning (CFL) highlights its practicality and advantages in distributed settings. New insights developed emphasize the importance of decentralized systems for the Metaverse’s evolution, with a blockchain-based QFL application demonstrated in a hybrid model. Our evaluation, implementation details and code are publicly available 1.	10.1109/TNSM.2025.3552307
Adam Kadi, Lyes Khoukhi, Jouni Viinikka, Pierre-Edouard Fabre	Adapting to the Evolution: Enhancing Intrusion Detection Through Machine Learning in the QUIC Protocol Era	2025	Early Access	Denial-of-service attack Protocols Machine learning algorithms	The advent of the QUIC protocol may herald a significant shift in the composition of online traffic in the years to come. The transport layer encryption of the QUIC protocol is one of its main evolutions, especially for metadata that was previously transmitted over TCP traffic without encryption. This new protocol has the potential to require significant alterations in future internet traffic analysis methods and impact network intrusion detection. On the other side, Machine learning has been used in several research projects to identify network intrusions, with positive outcomes. However, we must take into account new evolution of network traffic. In this paper, we propose a new approach that employs supervised machine learning algorithms to identify flows generated by bots interacting with a web server during a DDoS attack, focusing on the challenges posed by the QUIC protocol and its implications for effective intrusion detection and cybersecurity. Our contribution in this work is divided into three main parts: 1) A guided process with model architecture for emulating and collecting traffic that depict a range of situations our system may encounter; 2) an analysis module that consists on the creation of two labeled datasets, where observations represent the traffic flows detected in PCAP files. We studied the relevance of different features for these datasets, contributing to a thorough understanding of the quality of the data used; 3) a real world experimention for evaluating the effectiveness of several supervised machine learning algorithms on our datasets. This experimentation allows us to determine which algorithm provides the best prediction results.	10.1109/TNSM.2025.3540753
Yali Yuan, Yangyang Du, Guang Cheng	Class Incremental Website Fingerprinting Attack Based on Dynamic Expansion Architecture	2025	Early Access	Monitoring Fingerprint recognition Feature extraction	Encrypted traffic on anonymizing networks is still at risk of being exposed to the Website Fingerprinting (WF) attack. This attack can seriously threaten the online privacy of users of anonymity networks such as Tor. While deep-learning-based WF attacks achieve high accuracy in controlled experimental settings, they cannot continuously learn after deployment. In real-world environments, new websites are constantly emerging, requiring attackers to expand their monitoring scope continuously. This necessitates attack models capable of continuous learning and expanding classification capabilities. In this paper, we explore how attackers can leverage incremental class learning techniques to continuously learn new classes while retaining the ability to distinguish old ones. This approach mitigates the catastrophic forgetting problem in dynamic, open-world scenarios. We introduce a new WF attack, Class Incremental Fingerprinting (CIF), which employs a scalable architecture enabling Class Incremental Learning (CIL) with limited resources. We evaluate this attack in various scenarios, such as learning 100, 200, and 500 monitored website classes across 5 and 10 incremental tasks, achieving an average accuracy of 97.8% and above. Additionally, we assess the CIF attack’s effectiveness in open-world multi-classification scenarios and test it in few-shot settings using the proposed data augmentation method, Mixtam, achieving an average task accuracy of 87.6% and above with only 30 samples per class.	10.1109/TNSM.2025.3538895
Arad Kotzer, Bence Ladóczki, János Tapolcai, Ori Rottenstreich	Addressing Scalability Issues of Blockchains with Hypergraph Payment Networks	2025	Early Access	Blockchains Lightning Bitcoin	Payment channels are auspicious candidates in layer-2 solutions to reduce the number of on-chain transactions on traditional blockchains and increase transaction throughput. To construct payment channels, peers lock funds on 2-of-2 multisig addresses and open channels between one another to transact via instant peer-to-peer transactions. Transactions between peers without a direct channel are made possible by routing the payment over a series of adjacent channels. In certain cases, this can lead to relatively low transaction success rates and high transaction fees. In this work, we introduce pliability to constructing payment channels and graft edges with more than two endpoints into the payment graph. We refer to these constructions as hyperedges. We present hyperedge-based topologies to form hypergraphs and compare them to Bitcoin’s Lightning network and other state-of-the-art solutions. The results demonstrate that hyperedge-based implementations can both increase transaction success rate, in addition to decreasing the network cost by more than 50% compared to that of the Lightning Network.	10.1109/TNSM.2025.3542960
Jaidip Kotak, Idan Yankelev, Idan Bibi, Yuval Elovici, Asaf Shabtai	VPN-Encrypted Network Traffic Classification Using a Time-Series Approach	2025	Early Access	Telecommunication traffic Virtual private networks Cryptography	Network traffic classification provides value to organizations and Internet service providers (ISPs). The identification of applications or services from network traffic enables organizations to better manage their business, and ISPs to offer services to their users. Given the vast quantity of traffic flowing in and out of organizations, it is impractical to write manual signatures for traffic identification. The effectiveness of machine learning (ML) in the identification of applications or services from network traffic has been demonstrated. Even when network traffic is encrypted, ML algorithms achieve high accuracy in the task of traffic identification based on statistical information and the packets’ headers and payloads. However, existing approaches were shown to be ineffective for VPN-encrypted network traffic. In this study, we propose a novel time-series based approach for the identification of traffic/source applications on VPN-encrypted traffic. We also demonstrate the broad applicability of our proposed approach by evaluating its effectiveness on non-VPN traffic that is encrypted, and on IoT traffic.	10.1109/TNSM.2025.3543903
Xinrui Dong, Yingxu Lai, Xiao Zhang, Xinyu Xu	Counteracting New Attacks in CPS: A Few-shot Class-incremental Adaptation Strategy for Intrusion Detection System	2025	Early Access	Adaptation models Intrusion detection Training	The deep integration of physical devices and communication networks has increased the security risks of cyber-physical systems (CPSs) compared to traditional control systems. Deep learning-based intrusion detection systems (IDSs) play a crucial role in ensuring CPSs security. However, the existing IDSs often rely on known attack features, rendering them unable to withstand emerging new attacks arising from the dynamic evolution of intrusion behaviors. This paper aims to develop an IDSs with high adaptability and strong generalization capabilities, which is capable of rapidly adapting to new attack classes with only a few new samples. To achieve this objective, we propose CAT-IDS, a few-shot class-incremental adaptation strategy for an IDS to counteract new attacks on CPSs. We design a highly symmetric classifier structure for CAT-IDS that can flexibly adjust the classification space to adapt to new attacks. Furthermore, we calibrate the biased distribution formed by a few training samples through statistical feature transfer. In order to prevent the model from forgetting old attack information during the adaptation process, we devise hybrid features for attack detection. These features contain essential information for both old and new class classifications. We demonstrate the effectiveness of CAT-IDS through multiple experiments on three CPSs datasets. The results show that CAT-IDS achieves an average accuracy improvement of approximately 4. 5% compared to the state-of-the-art methods, demonstrating its superior ability to adapt to new attacks while maintaining high performance in classifying existing attacks.	10.1109/TNSM.2025.3543773
Amina Hentati, Amin Ebrahimzadeh, Roch H. Glitho, Fatna Belqasmi, Rabeb Mizouni	Deterministic and Dynamic Joint Placement and Scheduling of VNF-FGs for Remote Robotic Surgery	2025	Early Access	Surgery Haptic interfaces Robots	During a Remote Robotic Surgery (RRS) session, multimodal data traffic with different requirements is initiated. In order to achieve a cost-effective deployment of such a system, it is crucial to tailor resource allocation policies based on the different quality of service (QoS) requirements of each data traffic. In this paper, we focus on resource allocation in a 5G-enabled tactile Internet RRS system using network function virtualization (NFV). In particular, we investigate the joint placement and scheduling of Virtualized Network Functions (VNFs) in a RRS system under both deterministic and dynamic settings. An integer linear program (ILP) is used to formulate the problem. Due to its high computational complexity, we first propose an efficient greedy algorithm to solve the ILP under deterministic settings. Simulation results show that our proposed algorithm achieves near-optimal performance and outperforms the benchmark solutions in terms of cost and admission rate. It can reduce cost by up to 37% and improve admission rate by up to 34% while satisfying both latency and reliability constraints. Furthermore, our results show that modeling the multimodal data traffic by multiple VNF Forwarding Graphs (VNF-FGs) with different QoS requirements achieves a significant gain in terms of cost and acceptance rate compared to modeling it by a single VNF-FG with the most stringent requirements. We then considered a dynamic environment where latency variations and traffic arrivals may occur over time. Using the principles of optimal stopping theory, we propose an adaptive dynamic scheduler that is capable of triggering recalculations of the existing optimal solution based on the observed cumulative number of traffic arrivals and latency violations without the need for predictions. Our proposed optimal scheduler minimizes the migration cost compared to other schedulers.	10.1109/TNSM.2025.3539183
Jiajun Chen, Chunqiang Hu, Weihong Sheng, Ruinian Li, Ruifeng Zhao, Jiguo Yu	A Trust-Based Personalized Differential Privacy Guarantees for Online Social Networks	2025	Early Access	Privacy Protection Social networking (online)	Online social networks have emerged as a significant data source, but the extensive collection and utilization of personal information have given rise to profound concerns regarding privacy. From a legislative and policy perspective, and in alignment with the concept of privacy as control, users have the right to control their personal privacy information. However, users often encounter challenges in terms of understanding and effectively managing their privacy settings to align with their specific privacy requirements. To address this issue, in this paper, we incorporate the concept of trust and propose a trust-based personalized differential privacy model for online social networks, denoted as TPDP, which relies on a trusted central server to facilitate its operation. Specifically, when a user requests access to another user’s personal information, the TPDP mechanism provides a privacy response, where the privacy level is determined based on the direct and indirect trust values among users, calculated automatically by the trusted central server. Furthermore, the proposed TPDP model offers user-to-user personalized differential privacy protection from the perspectives of network structures, trust-related factors, and trust propagation patterns. Finally, we validate the model’s feasibility and assess the privacy-utility trade-off, as well as its robustness against attacks, through theoretical analysis and performance evaluation.	10.1109/TNSM.2025.3543844
Giampaolo Bovenzi, Francesco Cerasuolo, Domenico Ciuonzo, Davide Di Monda, Idio Guarino, Antonio Montieri, Valerio Persico, Antonio Pescapé	Mapping the Landscape of Generative AI in Network Monitoring and Management	2025	Early Access	Surveys Monitoring Artificial intelligence	Generative Artificial Intelligence (GenAI) models such as LLMs, GPTs, and Diffusion Models have recently gained widespread attention from both the research and the industrial communities. This survey explores their application in network monitoring and management, focusing on prominent use cases, as well as challenges and opportunities. We discuss how network traffic generation and classification, network intrusion detection, networked system log analysis, and network digital assistance can benefit from the use of GenAI models. Additionally, we provide an overview of the available GenAI models, datasets for largescale training phases, and platforms for the development of such models. Finally, we discuss research directions that potentially mitigate the roadblocks to the adoption of GenAI for network monitoring and management. Our investigation aims to map the current landscape and pave the way for future research in leveraging GenAI for network monitoring and management.	10.1109/TNSM.2025.3543022
Xingyu Yang, Jipeng Hou, Lei Xu, Liehuang Zhu	zkFabLedger: Enabling Privacy Preserving and Regulatory Compliance in Hyperledger Fabric	2025	Early Access	Blockchains Privacy Regulation	Preserving the privacy of transactions and ensuring the regulatory compliance of transactions are two important requirements for blockchain-based financial applications. However, these two requirements are somewhat contradictory. Techniques for protecting transaction privacy, such as data encryption and zero-knowledge proof, generally make it difficult to regulate and audit the transactions. In this paper, we propose a system named zkFabLedger which enhances both the privacy and the auditability of the classic permissioned blockchain platform Hyperledger Fabric. The proposed system utilizes commitments and non-interactive zero-knowledge proofs to hide the detailed information of transactions while enabling the endorsing peer nodes to verify the regulatory compliance of transactions. Transactions are recorded on table-structured ledgers, so that the regulator can perform complex auditing of transactions. Moreover, we utilize the ring signature scheme and the secret handshake protocol to ensure the anonymity of the transaction sender while enabling the regulator to trace the sender’s identity. Simulation results demonstrate that the proposed system can balance well between privacy, regulation and efficiency.	10.1109/TNSM.2024.3525045
Ahsan Raza Khan, Habib Ullah Manzoor, Rao Naveed Bin Rais, Sajjad Hussain, Lina Mohjazi, Muhammad Ali Imran, Ahmed Zoha	Semantic-Aware Federated Blockage Prediction (SFBP) in Vision-Aided Next-Generation Wireless Network	2025	Early Access	Sensors Wireless sensor networks Semantics	Predicting signal blockages in millimetre-wave and terahertz networks is essential for enabling proactive handover (PHO) and ensuring seamless connectivity. Existing approaches utilising deep learning, multi-modal vision and wireless sensing data primarily depend on centralised model training. Although these techniques are effective, they come with high communication costs, inefficient bandwidth usage, and latency issues, which restrict their real-time applicability. This paper proposes a Semantic-Aware Federated Blockage Prediction (SFBP) framework, leveraging the lightweight computer vision technique MobileNetV3 for edge-based semantic extraction, lowering communication and computation costs. Furthermore, we introduce a Similarity-Driven Federated Averaging (SD-FedAVG) mechanism to enhance the robustness of the model aggregation process, effectively mitigating the impact of noisy updates and adversarial attacks. Our proposed SFBP framework achieves 97.1% blockage prediction accuracy, closely matching centralised learning methods, while reducing communication costs by 88.75% compared to centralised learning and by 57.87% compared to FL without semantic extraction. Moreover, on-device inference reduces the latency by 23% compared to centralised learning and 18% compared to FL without semantic extraction, improving real-time decision-making for PHO. Additionally, the SD-FedAVG mechanism improves prediction accuracy under noisy conditions, directly impacting the PHO by reducing the handover failure rate by 7%.	10.1109/TNSM.2024.3525338