Immersive virtual reality (VR) stream is a major advancement in the context of next-generation multimedia services. However, its high bandwidth requirements bring new technical challenges to networked immersive service, while time-varying user interaction behaviors further increase complexity and introduce conflict of optimization objectives. To this end, we propose the Hierarchical Flatness-Bitrate Adaptive Transmission (HFB-AT) strategy that seeks to balance human visual field and media characteristics in VR streaming transmission systems. Inspired by the divergent characteristics of human vision, the new metric, i.e., flatness, is introduced to address the conflict between bitrate independence and spatial relevance of tiles. Then, the tile-based bitrate decision is abstracted as a two-level coupled decision of flatness and bitrate. We transform the adaptive decision process into a semi-Markov decision process, and the coupled two-level decision can be hierarchically inferred in different time intervals by comprehensively considering network conditions, user states, and historical information. To evaluate the direct impact of flatness on user experience, we design a new QoE evaluation model for VR streaming. Extensive experiments using public 4G/5G mobile network traces and real-world 5G measurements collected in Beijing compare HFB-AT with five representative heuristic-based, buffer-based, throughput-based, and learning-based baselines. The results demonstrated that HFB-AT effectively enhances user experience while avoiding extra bandwidth consumption, achieving a good balance between conservative and aggressive.

As Satellite Edge Computing (SEC) emerges as a computing backbone of 6G IoT, its feasibility is threat ened by frequent handovers that disrupt task continuity and cause deadline violations in AI-driven workloads. Conventional communication-centric handover strategies overlook the urgency of computational tasks, further exacerbating these issues. To address this gap, we propose the Computation Aware Conditional Handover Optimization (CACHO) framework, a non-intrusive solution fully compatible with 3GPP Conditional Handover (CHO). The framework integrates model partition driven delayed execution to enable seamless migration via inter-satellite tensor transmission, timeout-risk predictive proactive triggering to protect at-risk tasks from overloaded satellites, and task load-aware target selection to balance computational workloads and service popularity. These modules are orchestrated within standardized CHO workflows without altering the core logic, ensuring deployability in existing satellite network systems. Simulations based on real Starlink constellation data demonstrate that our approach reduces task timeout rates by 29.95–51.02% compared to the 3GPP standard CHO and several state-of the-art baselines, while maintaining handover success rates and minimizing extra handovers needed. Beyond performance gains, this work provides the first blueprint for embedding task life cycle awareness into handover workflows, bridging mobility management and computation guarantees, and advancing the feasibility of robust SEC for mission-critical 6G IoT applications.

Federated Learning (FL) faces challenges from client data heterogeneity and resource-constrained mobile devices, which can degrade model accuracy. Personalized Federated Learning (PFL) addresses this issue by adapting shared global knowledge to local data distributions. A promising approach in PFL is model decoupling, which separates the model into global and personalized parameters, raising the key question of which parameters should be personalized to balance global knowledge sharing and local adaptation. In this paper, we propose a Federated Optimal Brain Personalization (FedOBP) algorithm with a quantile-based thresholding mechanism and introduce an element-wise importance score. This score extends Optimal Brain Damage (OBD) pruning theory by incorporating a federated approximation of the first-order derivative in the Taylor expansion to evaluate the importance of each parameter for personalization. Moreover, we move the metric computation originally performed on clients to the server side, to alleviate the burden on resource-constrained mobile devices. To the best of our knowledge, this is the first work to bridge classical saliency-based pruning theory with federated parameter decoupling, providing a rigorous theoretical justification for selecting personalized parameters based on their sensitivity to local loss landscapes. Extensive experiments demonstrate that FedOBP outperforms state-of-the-art methods across diverse datasets and heterogeneity scenarios, while requiring personalization of only a very small number of personalized parameters.

360° video streaming in Space-Air-Ground Integrated Networks (SAGIN) faces a critical dilemma in the context of SAGIN's heterogeneous network topology and its dynamic propagation delays: the trade-off between field of view (FoV) prediction accuracy and buffer stability. Traditional adaptive bitrate (ABR) and routing strategies do not address this trade-off due to their reliance on stable latency assumptions and lack of coordination. Unlike them, we regulate streaming from the perspective of FoV prediction-to-playback (FPTP) delay, which inherently links routing-induced propagation latency with bitrate-driven buffer dynamics. This motivates the proposal of a novel dual-channel temporal-spatial integrated proximal policy optimization (DTS-PPO) which combines a spatial channel that adapts routing to rapid topology and link variations, and a temporal channel that regulates bitrate in response to evolving buffer and prediction trends. DTS-PPO extracts spatio-temporal state features via lightweight neural encoders, fuses them for global representation, and trains a joint routing–ABR policy with proximal policy optimization. This integration enables coordinated decisions that stabilize the FPTP delay under dynamic conditions. Extensive evaluations with real-world FoV traces demonstrate that DTS-PPO improves QoE by 21.3%–34.8% in diverse scenarios, significantly outperforming state-of-the-art ABR and routing baselines.

As one of the latest features of ultra-high-definition media services, high frame rate can significantly enhance perceptual quality, but also increases codec complexity in the transmission chain, leading to additional overhead. In this paper, we carry out comprehensive offline experiments in which the codec overhead (e.g., energy and delay) shows a linear or even quadratic increase trend with various frame rates, while correspondingly, when the frame rate increases to 75FPS, its bitrate is 24.2% lower than that under 15FPS for several scenarios. This illustrates that the overhead is more significant than the load from data traffic in the frame rate control problem. Thus, we propose a Bilateral Adaptive video Transmission framework that establishes Bilateral game-theoretic Control (BAT-BC) between sender and viewer. Through dynamically adjusting frame rate for sender and service payment for viewer, BAT-BC can flexibly adapt to the external environment such as computational state and scenario changes and it is expected to provide viewers with a smoother experience. Furthermore, we extend it to the scenario including concurrent multi-viewer and discuss the effects of grouping utility. Finally, we design a prototype system and the proposed solution is deployed on it to evaluate the performance. The frame drop rate is reduced by 61%, resulting in a 31% improvement in subjective QoE. The objective metric achieves the same level of actual experience as a fixed 60 FPS under dynamic environment.

Current congestion control algorithms often fall short when balancing multiple objectives, leading to inefficient competition and subsequently diminishes network efficiency. To address this challenge, we introduce HydraCC, an innovative multi-objective congestion control solution. Through theoretical derivation, we elucidate the impact of algorithm parameters on several performance metrics, such as QoS fairness, responsiveness, throughput, and window fluctuation. Our analysis reveals that these objectives often correlate or even compete with each other. Building on this insight, HydraCC is designed to expand a Pareto solution set from an initial solution. Specifically, using a carefully architected method based on Krylov subspace iteration, we efficiently explore the evolutionary direction of Pareto solutions. Crucially, each solution presents its own strengths and costs, thereby offering a broader space for trade-offs. Through rigorous real-world experiments and simulations, we demonstrate HydraCC's adaptability, and its superiority to existing state-of-the-art congestion control algorithms across various performance metrics. Moreover, it significantly mitigates the issues associated with multiple coexisting flows. When compared to contemporary leading algorithms such as Orca, Cubic, and PCC-Vivace, the convergence stability of HydraCC shows an improvement in the range of 47.89 similar to 77.21%.

With the proliferation of IoT devices, botnet-based distributed denial-of-service (DDoS) attacks have become increasingly severe, overwhelming system resources with large-scale malicious traffic and posing a serious threat to online service availability. Multi-agent Moving Target Defense (MTD) systems enhance attack resilience by dynamically reconfiguring proxy server clusters. However, identifying stealthy attackers within these systems remains a major challenge, particularly intelligent adversaries who can mimic legitimate user behavior or those who exploit proxy address leaks to facilitate indirect attacks. Existing traffic-based detection methods struggle to effectively identify such highly evasive threats due to their limited accuracy. Therefore, improving user behavior evaluation in MTD systems to efficiently detect and mitigate stealthy attackers is a critical research problem. To enhance detection accuracy, this paper proposes a DDoS attacker detection model for multi-agent MTD systems based on spatiotemporal feature fusion, leveraging both temporal and spatial dimensions to improve resistance against evasion strategies. In the spatial domain, we integrate user characteristics with proxy-side metrics, expanding detection scope beyond usercentric evaluations. In the temporal domain, we track anomalous patterns across multiple defense cycles, capturing behavioral inconsistencies over successive proxy reassignments. By jointly analyzing spatiotemporal features and iteratively refining trust scores, our model effectively identifies persistent malicious attackers, significantly improving detection capability against stealthy threats. Additionally, we design an optimized decisionmaking mechanism that accounts for the complexity and diversity of attack strategies, enhancing adaptability and generalization through refined feature selection and classification strategies. Experimental results demonstrate that our approach outperforms conventional detection methods in key metrics such as attack response time, false positive rate, and false negative rate.

Data centers extensively employ topologies such as CLOS, BCube, and Xpander to manage their large networks of devices. Data center networks (DCN), which rely on traditional topologies that are inherently static and inflexible to traffic pattern variations, face increasingly the adoption of dynamic topologies facilitated by multi-mode converters, providing a more adaptable solution without the need to expand the existing communication resources. However, fully utilizing the heterogeneous information of DCNs and achieving a network-wide joint optimization of converters remain significant challenges in algorithm design. This paper marks the first foray into using graph representation learning tools for DCN topology optimization and introduces MorpheusClos as an innovative solution. We present the novel DyTa block, which integrates graph transformers with causal convolution for extracting heterogeneous DCN graph representations, enabling joint optimization of network-wide converters from a graphical perspective, thereby improving network adaptability and performance under diverse traffic conditions. Extensive experiments and simulations reveal that MorpheusClos significantly outperforms existing topologies, reducing average path lengths by 29% and congestion probabilities by 57%.

This paper addresses the rate control issue for Video-on-Demand (VoD) services in Low Earth Orbit (LEO) satellite Internet. LEO systems employ long-distance Non-Orthogonal Multiple Access (NOMA), where the transmission rate of the last hop directly determines the Quality of Experience (QoE) levels for the VoD users and the satellite’s energy consumption. Our research identifies two primary issues: (i) determining the transmission rate to ensure high user QoE while minimizing energy consumption, and (ii) ensuring fairness among users within the satellite coverage area. To address these issues, we model the multi-user VoD viewing process as a Partially Observable Markov Process (POMDP) and describe the interactions among users using a cooperative coalition game framework. We propose Harmony, a distributed and dynamic improvement solution based on the Deep Deterministic Policy Gradient (DDPG) approach. Harmony intelligently determines each user’s transmission rate by combining feedback from user applications and MEC server metrics, ensuring superior QoE levels, energy efficiency, and fairness. The trained Harmony can be adapted to various Adaptive BitRate (ABR) algorithms, providing scalability and immediate applicability in existing LEO networks. It can also achieve improved performance in dynamic user environments. Simulation results demonstrate that Harmony improves energy efficiency and fairness, while maintaining high QoE levels and reducing MEC traffic overhead by 28.1% to 62.6%.

This study addresses the inefficiency of triggering vulnerability areas and low test case effectiveness in directed fuzzing by proposing an Exponential-weight algorithm for Exploration and Exploitation (Exp3) algorithm-based system. The system locates suspicious targets through function flow graph similarity analysis, constructs reward functions using code coverage metrics, and dynamically optimizes seed mutation via a multiarmed bandit model to reduce randomness. Evaluated on Bind9 and Dnsmasq with AFL, AFLGO, SelectFuzz, and two variants (Upper Confidence Bound (UCB) / Exp3-based), the Exp3-based tool demonstrated superior performance in Bind9 tests, achieving higher target/path coverage without complex confidence interval calculations. This strategy provides an innovative solution for directed fuzzing efficiency enhancement.

The metaverse is rapidly gaining momentum, thanks to its inherent capabilities to create an immersive virtual environment that runs in parallel to the physical world. At the same time, it raises new technical challenges due to its unique characteristics. On the one hand, metaverse applications are composed of multiple computing sub-tasks, and performing these sub-tasks sequentially can hinder computational efficiency. On the other hand, current centralized task offloading for metaverse applications is contrary to the core concept of Web 3.0. Moreover, fair incentives for all participants are not fully considered. To address these issues, a new computing paradigm for metaverse is required. Hence, we propose a Blockchain-enabled Intelligent Dispersed Computing Framework (BIDC). In this paper, we first design a two-layered architecture and model the sub-tasks as a directed acyclic graph (DAG) by utilizing their dependent relations. Inspired by the interconnection of blocks in blockchain, BIDC transforms the execution process of sub-tasks into a mining process, cleverly integrating task computation with mining. On this basis, Mining Mechanism and Main Chain Confirmation Mechanisms are presented, to ensure the efficiency of task offloading and the fairness of reward distribution. Then BIDC transforms the overhead time minimization problem into a multi-party mining problem. By leveraging Actor-Critic-based Multi-Agent Reinforcement Learning, every device can dynamically adjust its own mining strategy to achieve the lowest latency. At last, experimental results demonstrate BIDC’s reliability, scalability, and superior service quality compared to existing state-of-the-art solutions.

HTTP live streaming delivers dynamically video content with varying bitrates to accommodate the dynamic real-time bandwidth fluctuations while considering diverse user preferences and device capabilities. Existing flow control solutions do not provide support for new features such as multi-source content transmission. In this paper, we propose a distributed multi-source rate control optimization algorithm (DMRCA) that maximizes the overall network bandwidth utility and improves viewer Quality of Experience (QoE). First, we model the rate control problem as a dual-optimized multi-source and multi-rate problem. Then, we decompose the problem into sub-problems of source rate selection and user rate adaptation and we prove that solving the original problem is equivalent to solving these two sub-problems. Furthermore, we propose DMRCA as a fully distributed algorithm to solve these sub-problems and derive an optimal solution and we discuss DMRCA’s complexity and convergence. Finally, through a series of simulation tests, we demonstrate the superiority of our proposed algorithm compared to alternative state-of-the-art solutions.

Fairness-aware federated learning (FFL) plays a crucial role in mitigating bias against specific demographic groups (e.g., gender, race, occupation) during collaborative training. Along with the ever-emerging new attack paradigms like gradient leakage and model poisoning, the reliability of FFL also obtains lots of research attention. Either UAV nodes or FFL aggregators could be untrusted adversaries. Although multiple security mechanisms involving encryption, obfuscation, Byzantine-robustness, and detection have been proposed, concrete to UAV networks, the majority of existing solutions are unfeasible due to high heterogeneity and limited resources among participants. Hence, in this paper, we propose mutually reliable FFL (MR-FFL), a stratified community-based framework to facilitate privacy protection (FFL aggregator's reliability) and poisoning elimination (client nodes' reliability) jointly for FFL in heterogeneous UAV networks. We first divide UAV nodes into both peer communities (PC) and colleague communities (CC) according to cross-participant similarity and task-oriented fitness, respectively. Thus, the arbitrarily settled learning tasks following fair principles can be efficiently completed by fine-tuned colleague communities, even in the presence of a large degree of heterogeneity among peer communities. Then, we integrate community-specific differential privacy into the MR-FFL process, to achieve privacy amplification as well as efficient and personal collaborative training at the same time. More importantly, we proposed a community-based credit evaluation to resist poisoning attacks in heterogeneous environments. The results on several standard datasets also highlight the performance of MR-Fed in terms of fairness, accuracy, and integrity jointly.

In the burgeoning field of digital media, the popularity of live 360-degree video has witnessed an upward trajectory, underpinned by its immersive experience. In live 360-degree video streaming services, transcoding enormous video tiles and providing low-latency services for global viewers require significantly more computing and communication resources than traditional live services. Inspired by the huge crowdsourcing computational resources available at crowd, this paper designs CA-Live360, a Crowd-Assisted Live 360-degree video streaming solution, in which transcoding and delivery are supported by crowdsourcing devices. Due to the dynamic availability of resources, it is challenging to achieve low-latency live services, while guaranteeing fair transcoding. Therefore, firstly an innovative fairness-guaranteed transcoding task assignment scheme is proposed. This scheme employs a novel Fair Bandit algorithm based on the combinatorial multi-armed bandit approach to achieve low-latency transcoding while considering time-varying available resources and fairness constraints. Secondly, to reduce delivery latency, we design a transcoding-aware delivery scheme, in which a provider-requester matching algorithm is introduced to realize effective requester scheduling. Real-world trace-driven experiments demonstrate the improved performance of the proposed CA-Live360 solution in terms of system delay and fairness.

The rapid advancement in robotics technology has catalyzed the emergence of Internet of the Robotic Things crowdsourcing, a novel paradigm in the digital economy era. However, the escalating complexity of crowdsourcing tasks, coupled with the constrained resources of robot nodes and the diverse demands of stakeholders, has resulted in inefficiency, which poses a significant challenge to the burgeoning robot crowdsourcing market in IoT. To address these challenges, we develop a comprehensive analytical model that encapsulates the interests of both task scheduler nodes and robotic nodes within Internet of Things, which integrates various factors such as communication, computation, mobility, latency, and energy, thereby tailoring utility functions for different stakeholders. Secondly, we propose a Taskdriven Robotic Crowdsourcing strategy based on Hierarchical Game (TRC-HG), which conceptualizes TNs and RNs as rational entities with sequential actions. At the first layer, the interaction between participants is transformed into a Stackelberg game. This aids TNs in determining optimal pricing strategies while guiding RNs toward the most efficient task-completion strategies. Furthermore, we delve into the collaborative dynamics within the RNs, a cooperative coalition formation strategy at the second layer is established, which iteratively determines node responsibilities and coalition members. The Nash stability and optimality of coalition are ensured, thereby maximizing the utility of RNs. Finally, we validate the performance through a series of high-fidelity simulation experiments. These experiment results, benchmarked against classical methods, highlight significant improvements in terms of latency, energy consumption, and cooperative efficiency.

Multipath TCP (MPTCP) technology efficiently leverages multiple paths for data transmission, achieving commendable throughput performance. However, existing MPTCP scheduling algorithms fail to meet the diverse and evolving demands of modern applications, which often extend beyond throughput to include preferences for low latency or stability. To address this gap, we introduce FlexMRS, an intelligent multi-objective MPTCP scheduling algorithm. Unlike prevailing algorithms that rely on black-box machine learning models, FlexMRS innovatively utilizes the Lagrangian relaxation method to overcome the traditional limitations of balancing multiple objectives in machine learning.Furthermore, we introduce the Krylov subspace iteration to obtain obtain the local Pareto optimal solution set, thus make FlexMRS could effectively address the switching convergence problem of different application requirement models.

In the era of Industry 5.0, with the deep convergence of Industrial Internet of Things (IIoT) and 5G technology, stable transmission of massive data in heterogeneous networks becomes crucial. This is not only the key to improving the efficiency of human-machine collaboration, but also the basis for ensuring system continuity and reliability. The arrival of 5G has brought new challenges to the communication of IIoT in heterogeneous environments. Due to the inherent characteristics of wireless networks, such as random packet loss and network jitter, traditional transmission control schemes often fail to achieve optimal performance. In this paper we propose a novel transmission control algorithm, aBBR. It is an augmented algorithm based on BBRv3. aBBR dynamically adjusts the sending window size through real-time analysis to enhance the transmission performance in heterogeneous networks. Simulation results show that, compared to traditional algorithms, aBBR demonstrates the best comprehensive performance in terms of throughput, latency, and retransmission. When random packet loss exists in the link, aBBR improves the throughput by an average of 29.3% and decreases the retransmission rate by 18.5% while keeping the transmission delay at the same level as BBRv3.

Recently increasing real-time network applications have raised a new requirement for the future 6G network transmission, which desires the information-update packets arrive at the receiver as timely as possible. Multipath TCP (MPTCP) can be employed in the real-time update transmission system by making use of multiple paths so as to promote the information status update. However, current optimization solutions of MPTCP focus on traditional transmission performance such as throughput, delay, loss, jitter, etc. None of these works considers the freshness metric for MPTCP to satisfy the real-time demands. What's more, most works about transmission model in Software Defined Network (SDN) are based on packet-level analysis without considering flow-level situation. In order to fill these gaps, this paper proposes a novel freshness-aware age optimization solution for MPTCP (FAMT) over SDN. FAMT introduces a new four-tuples to describe the transmission state in the SDN architecture. Then FAMT classifies the whole transmission process into two phases: connection establish phase and data transmission phase. Based on the two phases, FAMT evaluates and defines the age of information for different situations. In addition, FAMT designs an innovative age-assisted fluid model of MPTCP which can realize the four goals effectively.

Multiaccess edge computing (MEC) has revolutionized the delivery of large-scale mobile multimedia services by endowing network edge with computing and caching capabilities. This not only relieves the load on core networks, but also significantly reduces data access latency. However, deploying edge data centers with a high density to accommodate the growing demand for multimedia services is not cost-effective. With the rapid development of the Internet of Things (IoT) industry, recent studies have shown that by allowing UAVs with integrated computing and communication to form a mobile device cloud (MDC) environment via UAV-to-UAV (U2U) communications in IoT networks, UAVs can play an important role in assisting cellular networks with multimedia delivery and providing excellent service for IoT devices on the ground. While a MDC environment composed of UAVs offers flexibility and cost-effectiveness, the challenge remains in allocating caching resources in a timely manner to meet the dynamic content demands. To address this challenge, we design a novel MDC-enabled Caching (MDC2) framework, which makes use of the available caching and U2U communication capabilities to enable any UAV to obtain dynamically content from other nearby UAVs via the IoT network. By modeling the dynamic network status as a fluid-based system, MDC2 employs a dynamic caching allocation algorithm to minimize both service latency and caching costs. Extensive experiments demonstrate that MDC2 outperforms a state-of-the-art MDC multimedia delivery approach by improving average cache utilization with over 40% and reducing average access latency with more than 25%.

As a key enabling technology in intelligent heterogeneous Internet of Things (IoT), edge caching provides important support for reducing core network load and improving network service efficiency, especially for high bandwidth demand services represented by multimedia applications. However, external time-varying information is hard to be obtained comprehensively in a complicated heterogeneous IoT environment. Meanwhile, there exists the substitutability of content (e.g., videos with different bitrates), which is difficult to make caching decisions online in real-time to achieve fast feedback with low latency and avoid useless deployment. To this end, this article designs a transcoding-enabled online cache scheme for IoT video service with cloud–edge–terminal collaboration. First, we design a variable bitrate video routing strategy to dynamically retrieve content from cloud/edge according to user demands. Furthermore, the video caching problem is considered as an online convex optimization problem to learn utility gradient and determine the optimal caching strategy in real-time without any prior information. On this basis, we extend the problem to elastic networks with dynamic available resources and prove the sublinear regret and sublinear constraint violation. Finally, we summarized five video request data sets and carried out differentiated multiple verifications based on different request habits and content requirements. Compared with the most advanced algorithms in terms of delay, we evaluated the performance advantages of the proposed scheme.