With the development of the Industrial Metaverse, industrial services have gradually evolved from a single processing paradigm centered on data collection and monitoring into complex service workflows featuring cyber–physical collaboration, multiple stages, and strong sequential dependencies, thereby imposing higher requirements on network-side resource orchestration and end-to-end performance assurance. As an important mechanism for describing multi-stage service processes and resource mapping relationships, service function chaining (SFC) provides a viable approach for modeling and deploying Industrial Metaverse services. However, in multi-domain heterogeneous network environments, Industrial Metaverse services exhibit characteristics such as dynamic arrivals, the coexistence of multiple service types and structures, and continuous operation, which pose challenges for SFC deployment in terms of cross-domain resource coordination, online decisio-nmaking, and multi-objective trade-offs. To address these issues, this paper investigates the SFC deployment problem in multi-domain Industrial Metaverse networks, formulates an optimization model that jointly considers domain selection and intradomain resource mapping, and proposes a hierarchical learning-driven SFC deployment method. By modeling decision-making at the domain and node levels separately, the proposed method achieves cross-domain global coordination and fine-grained intradomain resource allocation. Simulation results demonstrate that the proposed method outperforms baseline schemes in terms of service acceptance rate and end-to-end latency, and can effectively adapt to multi-service and highly dynamic operating environments in Industrial Metaverse scenarios.

Human activity recognition (HAR) is an evolving technique that offers innovative solutions across various domains, such as healthcare, sports training, and human-computer interactions. This paper addresses the novel challenge of video-based activity recognition, focusing on detecting and classifying athletes' actions to enable precision sports training. Conventional HAR methods based on direct video analysis incur excessive computational overhead and constrained applicability. In contrast, our novel transformer-based framework, namely RSFomer, converts videos into multivariate time series, and then detects and classifies the athletes' actions. However, sports videos often suffer from severe occlusion, which introduces significant noise to the converted time series and thus deteriorates recognition performance. To address this challenge, we implement several innovative strategies to improve the robustness of our framework. First, we propose a dual-scale filtering mechanism that leverages the unscented Kalman filter and kinematic constraints to reduce noise and outliers in the converted time series. Second, we incorporate the masking mechanism and temporal slicing mechanism to enhance the transformer's ability to handle anomalies and extract multi-scale features for accurate action recognition. We perform extensive evaluations on our Boxing dataset as well as the UEA and FineGym datasets. The results demonstrate that our RSFomer is effective, outperforming existing state-of-the-art methods with significant advantages.

ABSTRACT Software‐defined networking (SDN) is an essential trend in the future development of networks. Due to hardware limitations, its progress undoubtedly counts on distributed management with multiple controllers to oversee the global network. The multiple network controllers heavily rely on each other to provide network services. However, concerns regarding issues like commercial privacy leakage have given rise to a conflict between the imperative of privacy protection and the necessity of data sharing. This paper focuses on studying the conflict between privacy protection and collaborative sharing. By mapping blockchain nodes to corresponding controllers and deploying them independently, the trusted sharing of domain information is achieved. Leveraging 6G technologies, significant latency issues introduced by uplink and download transmissions are addressed. Abstract network data is encrypted using attribute‐based encryption, ensuring sensitive information protection while reducing blockchain overhead. Furthermore, we have incorporated the Provenance Data model and established an alliance blockchain system for cross‐SDN network operator management. Finally, a prototype system is implemented, and through rigorous testing, the system's functional effectiveness and high efficiency are demonstrated.

Specific Emitter Identification (SEI) is pivotal for ensuring the security of the Internet of Things (IoT). Traditional deep learning-based SEI techniques often falter in real-world applications, particularly when distinguishing between legitimate and rogue devices amid noisy conditions and low Signal-to-Noise Ratios (SNR). To surmount these challenges, we propose a novel open-set SEI (OS-SEI) strategy that utilizes a Metric-enhanced Denoising Auto-encoder (MeDAE) architecture. This advanced framework incorporates a deep residual shrinkage network, significantly augmenting the denoising autoencoder’s capability, thereby bolstering its resilience against noisy environments. Further, the integration of discriminative metrics, such as center loss, markedly enhances feature discrimination, resulting in heightened accuracy of device identification. Our comprehensive experimental assessments, conducted on an Automatic Dependent Surveillance-Broadcast (ADS-B) dataset, underscore the superiority of our proposed OS-SEI method over existing models. The findings confirm our approach’s enhanced robustness to noise and its superior accuracy in device identification within open-set scenarios.

Cooperative caching is regarded as one of the most promising technologies for vehicular networks, as it significantly reduces content delivery latency by prestoring popular content at roadside units (RSUs) closer to vehicle users (VUs). However, accurately predicting popular content and subsequently making caching decisions to enhance the Quality of Experience (QoE) remains a complex challenge, with varied content preferences and performance levels across VUs. To address this issue, we propose a fairness-aware collaborative edge caching system that formulates the factors influencing QoE as a multiagent Markov decision process, with the aim of maximizing the long-term system revenue. Furthermore, we design an activity-based VU clustering and weighted prediction model and propose a training framework based on asynchronous federated learning for global model updates. The trained model extracts key features and accurately predicts popular content from historical data. Additionally, we present an attention-enhanced multiagent discrete soft actor-critic algorithm to tackle the complex caching decision problem and promote collaboration among RSUs. Extensive simulation experiments validate the superiority of our solution. Compared with other benchmarks, the proposed method improves service fairness by 50% and increases average system revenue by 16.7%.

The medical literature and records on diabetes provide crucial resources for diabetes prevention and treatment. However, extracting entities from these textual diabetes data is crucial but challenging. Named entity recognition (NER) - an important corner-stone technology of natural language processing - has been studied well in the general medical field. However, there is still a lack of effective NER methods to handle diabetes data. Briefly, there are three challenges in the real world, including 1) the large volume of diabetes-related data to be processed, 2) the lack of labeled data, and 3) the high costs of manual labeling. To mitigate those challenges, this paper proposes a novel NER method based on semi-supervised learning, namely SNER, for diabetes data processing. It utilizes large amounts of unlabeled data to solve the problem of lack of labeled data. Specifically, it filters the predicted labels based on their confidence and uncertainty scores to reduce the noise entering the model and divide them into positive pseudo-labels and negative pseudo-labels. Also, it utilizes negative pseudo-labels reasonably to improve the training effect of pseudo-labels. Experiments on two public diabetes datasets show that SNER achieves the best performance compared with existing state-of-the-art models.

With the progressive evolution of the sixth-generation (6G) network, an array of diverse application tasks is experiencing a steady surge, consequently intensifying the computational pressure. However, even with highly optimized task offloading approaches, ensuring overall service quality for rapidly expanding network applications remains challenging due to hardware resource limitations. This paper proposes a deep reinforcement learning-based algorithm utilizing a multi-agent approach in the End-Edge-Cloud architecture for 6G networks. The offloading issue can be reformulated to a decentralized partially observable Markov decision process, which transfers the NP-hard problem. We design an efficient algorithm based on multi-agent deep deterministic policy gradient (MADDPG) to observe the states of user equipments (UEs), edge servers, and cloud servers, thereby reducing offloading delay and energy consumption. Numerical results demonstrate that our proposed algorithm demonstrates superior performance compared to conventional and state-of-the-art approaches.

Space-aerial-ground integrated networks (SAGIN) is gaining eye-catching attention in 6G research. Comparing with terrestrial networks, SAGIN guarantees to provide three-dimensional (3D), seamless connectivity, global coverage and high resource usage efficiency. In addition, network softwarization (NetSoft) is recognized as the crucial attribute of 6G networks. With softwarization, traditional dedicated hardware will be decoupled into software blocks and general-purpose hardware. Tailored service requests can be implemented in the forms of chained software blocks (also called as slices) and coexist on top of these general-purpose hardware. The softwarization scheme can enhance the resource utilization and service diversity. Though SAGIN and NetSoft are separately studied well, their joint research is still in its infancy. In this paper, we focus on the research of softwarized SAGIN and propose one novel resource allocation and orchestration framework, labeled as Stice-Soft-SAGIN. The goal of our Stice-Soft-SAGIN framework is to provide reliable and efficient slicing service in quasi-static state. When receiving one slicing service request, our Slice-Soft-SAGIN will conduct the first procedure of available resource checking. After successfully doing the resource checking, our Stice-Soft-SAGIN will turn to conducting the slicing resource allocation and orchestration from three ordered parts (terrestrial part, aerial part, and satellite part). Take note that resources considered in Stice-Soft-SAGIN belong to wireless (spectrum) and wired (computing and storage) types. In order to validate the Stice-Soft-SAGIN, we conduct the evaluation in the simulation form. Evaluation results are illustrated and analyzed.

To alleviate the occurrence of flow table overflow in SDN network, this paper studies the flow table overflow mitigation strategy based on network flow path optimization. This paper first describes the phenomenon of network flow table overflow and its causes, and analyzes the existing solutions. Then, based on the centralized control function of the software-defined network controller, a Flow Table Resource-based Routing Algorithm (FTR) is proposed. The algorithm mainly includes three parts: calculating the load factor of the switch, evaluating the balance degree of the network flow table, and calculating the link weight and routing. Finally, the effectiveness of the improvement measures proposed in this paper is verified by simulation.

In this paper, we research the survivable service function chaining (SFC) mapping in network function virtualization (NFV)-enabled 5G networks. Previous SFC mapping studies were conducted on the basis of the assumption of no physical element failure. To tackle this issue, we propose one SFC mapping framework to deal with the single physical node failure. The proposed framework is labeled as Sur-SFC-E. When processing one virtual network service, usually modeled as a SFC, our Sur-SFC-E framework enables to map it successfully in the first place. When one single physical node comes into failure and certain virtual node is running on top of it, our Sur-SFC-E will do the recovery and make the executed service survivable. In order to highlight the Sur-SFC-E advantage, we do the simulation work. The counterpart without considering recovering from the single node failure is selected for performance comparison. We plot and discuss the success ratio and resource utilization results.

Ad-hoc networks belong to the distributed networks that do not depend on the fixed infrastructure and can dynamically change the network topology and structure. Therefore, it is of great value to promote the ad-hoc networks to the actual application. However, the openness of the ad-hoc networks makes them vulnerable to malicious attacks. Till now, multiple researchers have carried out a lot of work on solving ad-hoc security issues. This paper discusses these security threats and solutions from different aspects in the first place. Then, this paper analyzes the existing security solutions. Next, this paper summarizes the advantages and disadvantages. Finally, this paper proposes future research directions.

Virtual network embedding (VNE) refers to the resource allocation problem for network virtualization. Since its inception, multiple mapping algorithms have been proposed for embedding virtual networks (VNs) effectively and efficiently. However, prior mapping algorithms mostly complete the VN embedding in two separated stages: first node embedding and subsequent link embedding. Certain mapping algorithms embed the VN in one stage by using mixed integer linear programming method or subgraph isomorphism approach, involving high embedding completion time. Meanwhile, prior researchers conduct the VN embedding, on the basis of one underlying substrate network (SN). While in future VNE application, each VN must be mapped among multiple geographically distributed SNs. On above backgrounds, we propose a location aware and node ranking value driven embedding algorithm, labeled as LANRVD. The LANRVD enables to conduct the embedding in two coordinated embedding stages within polynomial time. In addition, the LANRVD embeds the VN among multiple geographically distributed SNs. Numerical results reveal that the LANRVD significantly improves VN acceptance ratio by 10% over existing typical two-separated-stages algorithms.

In the area of network virtualisation, virtual network embedding (VNE) refers to the resource allocation problem. In the literature, researchers have proposed multiple VNE algorithms. These algorithms have the goal of accommodating as many requested virtual networks (VNs) as possible. However, most of prior embedding algorithms belong to the two‐stage (separated node and link embeddings) mapping algorithm category. Certain embedding algorithms embed each VN in one mapping stage by using mixed integer linear programming approach or graph theory, having very high computation time. There is a lack of heuristic algorithms, enabling to embed nodes and links per VN in one mapping stage. In addition, each requested VN embedding needs to be completed in polynomial time so as to be promoted to future dynamic VN service application and real‐time VNs embedding. Based on these backgrounds, the authors propose a novel real‐time and one‐stage heuristic mapping algorithm (VNE‐RTOS). Numerical evaluations are conducted to strengthen that VNE‐RTOS earns more embedding revenues by 8% over typical two‐stage heuristic embedding algorithms (e.g. VNE‐TAGRD) while achieving the same substrate resource utilisation.

Virtual network embedding (VNE) refers to the resource allocation problem in network virtualization (NV). In the literature, researchers have proposed multiple VNE algorithms. These algorithms aim at embedding more and more requested virtual networks (VNs) onto the underlying networks and maximizing embedding revenues. Prior VNE algorithms mostly belong to the two-stage (separated node and link embedding) mapping algorithm category. Some other VNE algorithms embed each VN in one stage by using mixed integer linear programming (MILP) approach. There is a lack of one-stage heuristic algorithm, enabling to embed nodes and links in one mapping stage. In addition, each requested VN needs to be mapped in polynomial time so as to be promoted to future dynamic VN service application and real-time VNs embedding. Therefore, we propose a real-time and one-stage heuristic mapping algorithm (VNE-RTOS). Numerical simulations are conducted to validate that our VNE-RTOS earns more embedding revenues by approximately 3.4% over typical two-stage heuristic embedding algorithms (e.g. GRD-VNE) while achieving the same substrate resource utilization.