With the decreasing cost of satellite launch, Low Earth Orbit (LEO) satellite constellations have become an important part to complement the terrestrial communication systems for seamless coverage, limited latency, and high throughput. Meanwhile, the developing hardware computation capacity and Inter-Satellite Links (ISLs) have enabled LEO satellites to collaboratively conduct the onboard data processing, which is significantly important for the on-orbit Earth observation, environmental monitoring, and space exploration. However, the highly dynamic topology and heterogeneous resource distribution of LEO networks lead to traditional terrestrial scheduling mechanisms being ineffective. To address these challenges, this paper proposes a collaborative computation offloading framework based on the Software Defined Networking (SDN) technique for heterogeneous LEO satellite networks. The logically centralized controllers obtain global network states (e.g., satellite computing load and ISL conditions) to enable flexible-adaptive task scheduling. A Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm is proposed to decide the task splitting ratio, task processing satellites, and result aggregation satellite under dynamic network conditions. The simulation results demonstrate that the proposed MADDPG strategy significantly reduces the average task completion latency and improves the task success rate compared to benchmarks.

Implementing federated learning (FL) over wireless networks faces critical research challenges, such as high communication costs, inevitable communication latency, and significant energy consumption for model transmission and training, primarily caused by device heterogeneity and unpredictable dynamic channel conditions. This paper proposes Graph-based Resource Optimization with Compression for FL (GROC-FL), a unified framework that jointly coordinates wireless resource allocation and collaborative model compression. By leveraging Graph Neural Networks (GNNs) to model wireless topology and Deep Reinforcement Learning (DRL) to optimize communication and computation resources together with a globally consistent sparse update mechanism, GROC-FL minimizes the overall energy consumption of clients over wireless networks. In order to address the intrinsic topology dependence of wireless FL, we develop a graph-augmented DRL agent based on a graph convolutional network (GCN) that captures resource competition and network topology. We further develop a collaborative model compression module, termed Federated Parameter Negotiation (FPN), which enables clients to negotiate a global sparse mask and further reduce energy consumption during FL training. Experimental results demonstrate that GROC-FL outperforms the baselines in energy consumption, training performance, and client fairness.

The rapid iteration of artificial intelligence (AI) and 5G technologies has created essential foundations for Metaverse, easing bottlenecks in content generation, real-time interaction, and large-scale deployment. Integrating unmanned aerial vehicle (UAV) with semantic communication (SemCom) further provides a lightweight sensing and uplink solution, but semantic compression and bias may degrade fidelity. Generative AI (GAI), with powerful contextual modeling and generation capabilities, helps mitigate these issues. Based on this, this paper proposes a GAI-aided SemCom architecture and evaluates the feasibility of applying it to low-altitude internet of things (IoT) networks. We build evaluation models for representative image compression and GAI-aided SemCom techniques, replacing traditional task assumptions with realistic cases. Then, taking UAV networks as an example, we design a path planning and two task allocation schemes. Specifically, the UAV path planning strategy together with a greedy task allocation scheme improves the Metaverse update frequency and ensures the data freshness, while a parametrized deep Q-network (PDQN) task allocation scheme jointly optimizes Metaverse fidelity and freshness through mixed action selection, i.e., preprocessing methods (discrete) and compression size (continuous). Extensive analysis and numerical results corroborate that GAI-aided SemCom technology has practical significance in low-altitude IoT networks. Through the collaboration of multiple data processing schemes and the rational resource allocation, the fidelity and freshness of Metaverse can be greatly improved.

Mars exploration holds the potential to uncover the origin of life, which requires a reliable and sustainable electrical power supply to support long-duration missions. Wireless Power Transfer (WPT) based on Space Solar Power Satellites (SSPSs) has emerged as a promising option due to its continuous 24 × 7 availability and renewable nature. In particular, Laser-based WPT (LWPT) is especially suitable for space energy transmission due to its narrow beam width and high directivity. However, the extremely long propagation distance, high dynamics, and harsh conditions such as atmospheric attenuation, dust storms, and plasma effects degrade the transmission efficiency. To improve the robustness of energy delivery under such uncertain conditions, we incorporate Intelligent Reflecting Surfaces (IRS) to reconfigure the wireless propagation environment. Specifically, IRSs are exploited to provide alternative reflective transmission paths and enable adaptive wavefront control at the Mars rover, thereby enhancing the reliability and efficiency of power transfer when the direct link is partially blocked or severely attenuated. Additionally, we consider the presence of a malicious rover attempting to steal energy or disrupt legitimate charging. To ensure secure charging, we propose an IRS-assisted Challenge-Response Physical-Layer Authentication (CR-PLA) scheme. The False Alarm (FA) and Miss Detection (MD) probabilities are adopted as key performance metrics to quantify the authentication reliability. Finally, we formulate an optimization problem to maximize the harvested energy at the legitimate rover by jointly optimizing satellite selection, active transmit beamforming, and the passive IRS reflection, subject to the FA and MD probabilities. To address this intractable non-convex problem, we propose an Alternating Optimization (AO) algorithm to solve it iteratively. Numerical results demonstrate that, compared with benchmark methods such as Successive Convex Approximation (SCA) and greedy approach, the proposed method significantly enhances the harvested energy while effectively reducing MD probability.

The autonomous driving, which integrates wireless communication, intelligent computing and environmental perception, not only improves traffic safety and reduces vehicle accidents, but also alleviates traffic congestion by optimizing traffic flow and brings comfortable and convenient travel experience to passengers. However, current autonomous driving technology is generally at the L3-L4 levels and faces many challenges, such as semantic segmentation. Semantic segmentation enables vehicles to correctly distinguish the surrounding environment, such as roads, vehicles, pedestrians, etc. It can assist drivers in perceiving the surrounding environment well and making correct decisions, improving driving safety. However, current semantic segmentation work mostly focuses on improving recognition accuracy and neglects inference speed. Slightly higher latency can easily prevent vehicles from making timely and correct decisions, leading to accidents such as car crashes. Therefore, we propose a Multi-Level Real-time Fusion Semantic Segmentation Network (MLRFNet) that improves inference speed while ensuring high semantic segmentation accuracy for autonomous driving. The MLRFNet utilizes two lightweight branches, achieving effectively extract RGB and depth features with low computational cost. In addition, the Feature Fusion Module (FFM) aggregates complementary features from them, while the Cross-Level Refine Module (CRM) merges high-level semantic features and low-level spatial information. Extensive experiments demonstrate that MLRFNet significantly improves the inference speed while ensuring high accuracy. On the Cityscapes validation set, MLRFNet achieves 251.8 FPS and 71.4% mIoU for 512 × 1024 images inputs.

As an emerging application of the Internet of Things (IoT), autonomous vehicles (AVs) has attracted widespread attention from scholars. Accurate vehicle positioning is crucial for AVs to navigate safely and efficiently. Among the key components of positioning systems, odometry plays a vital role in tracking the vehicle’s movement, especially when GPS is unavailable. Traditional single-modal odometry methods, which rely solely on either LiDAR or cameras, often experience limited accuracy in challenging environmental or weather conditions. Therefore, some researchers proposed an idea of combining information from both LiDAR and cameras, called LiDAR-camera odometry (LCO). However, the existing LCO methods face issues like insufficient data integration or complexity in system structure. To address these challenges, we propose a novel LCO method named EVP-LCO, which enhances the sparse features from LiDAR by pseudo-LiDAR. In our method, we use a data augmentation module to enrich the details of LiDAR point cloud. Furthermore, we devise a feature regrouping strategy in two-step Levenberg-Marquardt (LM) optimization process to estimate accurate pose and reconstruct colorful global map. The results on the KITTI Odometry dataset show that EVP-LCO significantly improves vehicle positioning accuracy.

Intelligent vehicles are becoming an essential means of transportation for users, which provide them with high-quality services and convenient travel experiences. It should be noted that while the number of intelligent vehicles is growing rapidly, the incidence rate of traffic accidents is also increasing synchronously. Thus, some researchers try to reduce the risks through trajectory prediction. The current trajectory prediction schemes can be divided into single-vehicle-based scheme and vehicle-road collaboration scheme. However, single-vehicle-based scheme is lack in the states of surrounding vehicles and environment. Vehicle-road collaboration scheme is unable to overcome the inevitably limited field of view. In addition, present trajectory prediction schemes are limited by the computing resources of vehicles. Digital twin has become a hot topic due to its high computing speed and adequate environment information based on its cloud platform. Combining digital twin and trajectory prediction can not only broaden the vehicle’s field of view but also alleviate vehicle computing resources. For this, we first propose an adaptive trajectory prediction scheme based on digital twin platform. The whole computational process is placed in the cloud to alleviate the limited computing power while improving the computing speed. Vehicles can obtain the information of surrounding environment and the prediction results in real time, which improves the accuracy of the data used for prediction. Moreover, we utilize multichannel attention mechanism to learn the trajectory’s relevance weights adaptively, which improves the processing speed of scheme. The experimental results show that the performance of our scheme is better than others, so that drivers can adjust their routes to avoid collisions.

With the advantages of seamless coverage and ubiquitous connections, Non-Terrestrial Networks (NTNs) composed of Low Earth Orbit (LEO) satellites and Unmanned Aerial Vehicles (UAVs) can provide content caching services for future Connected Automated Vehicles (CAVs) to satisfy onboard collaborative viewing, traffic sensing, and metaverse entertainments in remote areas. However, the heterogeneous caching hardware, communication environments, and frequent network dynamics make the optimization of content caching policy highly complicated. Firstly, considering all LEO satellites as caching satellites can lead to content duplication and radio interference, causing storage waste and NTN transmission quality deterioration. Secondly, how to provide customized QoS by intra-layer and inter-layer cooperative caching in such complicated environments remains an open issue. Thus, we propose a Delay-Motivated Ant Colony Optimization (DM-ACO) scheme to select caching LEO satellites with reduced system propagation delay. Then, the Multi-Agent Deep Reinforcement Learning-based Hierarchical Caching and Asynchronous Updating (MADRL-HCAU) strategy is designed to manage the caching capacity of LEO satellites and UAVs, providing customized services for CAVs and dispensing the peak traffic. Simulation results illustrate that the proposed scheme can not only effectively accelerate the caching refreshing and content downloading process but also significantly reduce the packet drop and improve the cache hit ratio.

The development of 6G should enable users in remote and harsh areas to enjoy computation-intensive services including metaverse entertainment, intelligent transportation, and immersive communications. Low Earth Orbit (LEO) satellite constellations widely constructed in recent years have been recognized as an efficient solution to complement the terrestrial infrastructure with seamless coverage and decreasing expenses for both communication and computation services. However, the widely studied Federated Reinforcement Learning (FRL) based task offloading strategies neglect the potential trust concerns like malicious satellites and buffer pollution, while 6G service providers may rent the LEO satellites belonging to different companies to minimize the expense. To address these issues, blockchain has been considered in the Zero Trust (ZT) scenario, with the group consensus mechanism through the smart contract. Moreover, we propose a Constrained Correction Voting Mechanism (CCVM) to give punishing correction to the aggregation weight of malicious voting satellites. Furthermore, a Cold Start Reputation Aggregation (CSRA) scheme is adopted to first severely degrade and then gradually recover the weight of Federated Learning (FL) sub-models trained by malicious satellites. Thus, the Blockchain-enabled Cold Start Aggregation FRL (BCSA-FRL) scheme is proposed to make effective and secure offloading decisions in the ZT LEO satellite Networks. The numerical results illustrate the advantages of our proposal.

Roadside units (RSUs) play a vital role in intelligent transportation systems (ITS), working as critical elements in delivering superior Internet of Vehicles (IoV) services. A large service coverage and fast accident information diffusion RSU deployment solution can reliably ensure the ITS’ quality of service. Simultaneously, with the development of the city and the ITS, changes in traffic flow lead to RSU load imbalance, which will reduce the benefit of the original RSU deployment, and it is necessary to adjust RSU locations with minimal cost. Besides, due to the high visibility of the ITS, RSUs are highly susceptible to external attacks, which is commonly overlooked in existing RSU deployment work. Specifically, Sybil attack is one of the most dangerous attacks against ITS, and it can reshape the network state by forging multiple identities, interfering with risk sensing, etc. Motivated by these, we respectively propose the PSO-meme joint heuristic deployment algorithm (PJHDA) and the heuristic RSU multi-objective adaptation adjustment algorithm (HRMA3) to carry out deployment and adaptation adjustment of the city’s RSUs, taking into account the constraint of Sybil attack detection. Numerical results demonstrate that the multi-attribute performance of PJHDA is superior to the existing schemes. Compared with benchmark schemes, the HRMA3 excels in achieving advanced service coverage and load balancing while controlling costs, and both proposed schemes exhibit higher Sybil attack detection rate.

Industry 4.0 generates a huge volume of data, where Federated Learning (FL) can be utilized to mine the data in a privacy-preserving manner. However, traditional FL in privacy-preserving is not sufficient, the uploaded local model gradients can be intercepted by external Eavesdroppers (Eve), with more enough of which the users’ raw data can be inferred, leading to privacy leakage. At the same time, the future 6G accommodating more devices, makes privacy concerns sharper. To tackle privacy issues in FL, in this paper, we propose a Model-Safeguarded FL framework based on Intelligent Reflecting Surface (IRS) (MSFL) where Non-Orthogonal Multiple Access (NOMA) is introduced to enable multiple devices access. Specifically, IRS is deployed between Base Station (BS) and participants, improving the wireless environment and preventing Eve from eavesdropping. The Deep Deterministic Policy Gradient (DDPG)-Optimized Power and Phase (DOPP) algorithm is proposed to jointly optimize transmission power at participants and IRS phase shift to maximize the minimum confidentiality capacity. Extensive results demonstrate that the maximum confidentiality capacity of our MSFL scheme is up to 1.7 bps/Hz at a transmission rate of 30 dBW, which is approximately 300% more than that of the Block Coordinate Ascent Method (BCAM) and Artificial Noise (AN).

The development of Autonomous Vehicles (AVs) provides users with high-quality services and convenient travel experiences. As one of the most important functions in the automotive field, mobile positioning has attracted widespread attention from scholars. However, using a single-modal sensor (LiDAR or camera) poses challenges for precise localization due to their measurement flaws. Therefore, some scholars have proposed Visual-LiDAR Odometry (VLO). Nevertheless, most of the existing VLO solely use a single-modal sensor as their main framework and utilize another sensor for optimization, which does not fully leverage the complementary behavior of sensors in different environments. Thus, this paper presents a novel LiDAR-camera fusion method for improving the odometry estimation. Firstly, we employ a depth completion network to convert the image into pseudo-LiDAR to compensate for the missing depth values in the LiDAR point clouds. Then, we adopt Bayesian inference to enhance the robustness of the fusion method in different environments. Finally, evaluations on the public KITTI odometry show that the proposed method outperforms several state-of-the-art methods.

The development of 5G-vehicle-to-everything (5G-V2X) technology makes vehicle-road-cloud collaboration possible. Vehicles and roads transmit sensor data to the cloud via 5G-V2X technology and then the cloud sends the data to the target vehicle. The target vehicle utilizes dynamic environmental data from surrounding vehicles and roadside units to predict the driving trajectory of surrounding vehicles in order to ensure its own safety. However, many existing trajectory prediction schemes are based on incomplete single-vehicle perception and ignore surrounding road conditions, which will greatly limit their value in real-world scenarios. Therefore, this paper proposes a flexible multi-channel vehicle trajectory prediction scheme based on vehicle-road collaboration. Specifically, we first design a flexible multi-channel vehicle trajectory prediction scheme that can extract different vehicle and map features from various information sources. Then, we use the Transformer model to generate predicted trajectories of surrounding vehicles by fusing features from different sources, and achieve parallel computing effects. The most popular dataset, INTERACTION, is used to evaluate the proposed scheme. The results show that our scheme is robust across different scenarios and possesses better accuracy.

As a critical element in intelligent transportation systems (ITS), roadside units (RSUs) are pivotal in delivering superior Internet of Vehicles (IoV) services encompassing intelligent traffic management, accident prevention, and emergency rescue. Considering the high deployment and maintenance costs of RSUs, many studies focus on the efficient RSU deployment issues. However, due to the high visibility of the ITS system, RSUs are highly susceptible to external attacks, which is commonly overlooked in existing RSU deployment researches. Specifically, the Sybil attack is one of the most dangerous attacks against ITS, it can reshape the network state by forging multiple identities, interfering with the operator’s reputation assessment or causing severe DDoS. Therefore, we propose a joint heuristic scheme that combines the advantages of particle swarm optimization and double local-search memetic algorithm to solve the city RSU deployment problem in Sybil attack environments. It can find solutions with higher fitness values and guarantees that the IoV has the capability to detect Sybil attacks. Numerical results show that our proposed scheme not only outperforms other traditional solutions regarding signal validity coverage, overlap rate, and initial propagation speed of accident information, but also performs satisfactorily in Sybil attack detection.

With the advantages of seamless coverage and ubiq-uitous connections, Non-Terrestrial Networks (NTNs) composed of Low Earth Orbit (LEO) satellites and Unmanned Aerial Vehicles (UAVs) can provide content caching services to reduce End-to-End (E2E) delay and alleviate the network traffic for future 6G applications including autonomous driving, eHealth, and metaverse. However, the ultra-density of LEO satellites complicates the selection of caching nodes, while the heteroge-neous caching hardware and communication environments make optimization of content deployment highly difficult. To address these issues, we propose an intelligent hierarchical caching and asynchronous updating scheme. Specifically, a Delay-Motivated Ant Colony Optimization (DM-ACO) scheme is first adopted to select the caching LEO satellites to reduce the system propagation delay. Then, the Multi-Agent Reinforcement Learning-based Hi-erarchical Caching and Asynchronous Updating (MARL-HCAU) strategy is proposed to meet caching service demands. Simulation results illustrate that compared with the benchmarks, the overall cache hit ratio increases by 14.2 % with the reduced packet drop rate and transmission delay by 8.78 % and 0.94s, respectively.

As an essential 6G component, the Low Earth Orbit (LEO) satellite communication has aroused increasing attentions from academia and industry to provide seamless and highly-efficient networking services. However, existing routing strategies are primarily designed for terrestrial networks or small-scale satellite networks, making it inapplicable to future LEO satellite constellations of ultra density, high dynamics, and large scale. Moreover, since Free Space Optical (FSO) communications have been expected for Inter-satellite Links (ISLs) and the number of constructed FSO ISLs depends on the Acquisition, Pointing, and Tracking (APT) terminals and geometric visibilities, the routing algorithm needs to be adaptive. To address these issues, this paper considers the dual-layer network architecture composed of Medium Earth Orbit (MEO) satellites and LEO satellites, where the regional network division is adopted for the LEO satellite layer to alleviate the complexity and improve the routing efficiency. Then, a multi-objective reinforcement learning-based routing strategy with local information considered is proposed to meet the differentiated Quality of Service (QoS) requirements of diversified terrestrial applications. A cooperative mechanism is also designed to address the conflicts caused by the routing design for different applications. The simulation results demonstrate the proposal is applicable to varying numbers of APT terminals and outperforms benchmark algorithms in terms of diversified QoS metrics.

Behavior-oriented decision-makers are critical components in generating intelligent decisions for user virtual interactions in metaverse. In this work, we study the efficiency and security of behavior-oriented decision-makers in metaverse from perspective of autonomous driving (AD), where modeling human uncertain driving behaviors is the key factor of their performance. We first explore the ability of different deep-neural-network-based decision-makers used in deep reinforcement learning for efficient autonomous vehicle control, and then we propose a novel neural backdoor attack against them using spatiotemporal driving behaviors, rather than an immediate state. With our attack, the adversary acts as a normal driver and can trigger attacks by driving his vehicle following specific spatiotemporal behaviors. Extensive experiments show that our proposed backdoor attack can achieve high stealthiness and effectiveness (less than 1% clean performance variance rate and more than 98% attack success rate) on behavior-oriented decision-makers, and is sustainable against existing advanced defenses.

High accuracy and simultaneous positioning is an essential demand in future Intelligent Transportation Systems (ITS), while the mobility and dynamics of vehicles place great challenges. Single Base Station (BS) positioning has become popular for its fast speed, high convenience, and low cost. With the construction of 5G, the wide bandwidth and high separation capability of millimeter Wave (mmWave) bring more possibilities for vehicle positioning via single BS. However, mmWave signals have high distance attenuation and are easily blocked by obstacles. In urban scenarios, the prevalent None-Line-of-Sight (NLoS) situations have severe impacts on positioning accuracy. The multipath effects, Doppler effects, and tracking lags further degrade the performance. To address these issues, we introduce the Intelligent Reflecting Surface (IRS) to single BS vehicle positioning for beyond Line-of-Sight (LoS) communications. We study the advantages of IRS in urban ITS to alleviate the multipath effects, Doppler effects, and tracking delay. To realize the real-time target tracking for IRS, the Unscented Kalman Filter (UKF) is adopted, for which stable communications between the BS and moving vehicle can be maintained. Simulation results show that the utilization of IRS can significantly improve the positioning accuracy and the adoption of UKF further enhances the performance.

The Internet of Vehicles (IoV) is a large-scale interactive network that operates in a dynamic and changeable environment, encompassing diverse types of private information. In recent years, safeguarding vehicle location privacy in IoV has been a topic of concern. However, the independent location privacy protection mechanisms cannot consistently meet the rigorous security requirements of various IoV scenarios, which will pose a significant threat to the location privacy of IoV users. In contrast, risk sensing as a preventive security strategy needs lower computing costs and is more suitable for the intricacies of complex city environments. Unfortunately, the existing works lack that combination of risk assessment with trust assessment to conduct a comprehensive study on location privacy risk sensing. Therefore, considering the city vehicles’ spatial clustering phenomenon and the strong regularity of traffic flow, we propose a risk-sensing approach to vehicle location privacy based on the continuous adaptive risk and trust assessment strategy. This approach employs the Ripley method to analyze space clustering characteristics and combines the traffic flow prediction model to establish the risk assessment scheme. Furthermore, to enable our risk-sensing approach to have historical memory that can identify and continuously track malicious users, we incorporate a penalty factor into the trust assessment scheme that updates in a time iterative format. Extensive numerical results demonstrate the adaptability and reliability of our proposed risk-sensing approach to vehicle location privacy.

The Internet of Things (IoT) Network brings more possibilities and connections to indoor scenarios. With the growing number of mobile devices and online services, indoor wireless networks are required to achieve higher speed and capacity for carrying the exponentially increasing amount of data. The Tera-hertz (THz) communication, with its large spectrum resources, is considered to be the main technology to achieve ultra-high-speed wireless communications in the 6G era. However, THz signals are severely attenuated over distances and the transmission of narrow-beam signals in THz is severely affected by the None-Line-of-Sight (NLoS) environment. To tackle this indoor transmission issue, Intelligent Reflecting Surface (IRS) attracts attention with its beyond Line-of-Sight (LoS) ability and easy deployment. However, an IRS can only serve a limited number of users and requires proper allocation due to the constraints on hardware. The transmission performance is also severely affected by the interference among sub-channels. In this paper, we study IRS-aided data offloading for indoor THz networks. Considering the user interference, transmission conditions, and diversified tasks, we optimize the allocation of limited IRS resource as well as partition planning. A dynamic algorithm based on combina-torial optimization is proposed. Simulation results reveal that our proposed approach can significantly improve the offloading performance.

The rapid development of 5G and battery has enabled the electricity-driven intelligent connected vehicles to become the focus of current automobile industry. With automobiles growing intelligent, convenient, and entertaining, the computation tasks generated by various vehicle-integrated applications significantly increase. Cloud servers far away from the vehicles cannot complete the users' tasks in time, while the energy and computing resources on current Electric Vehicles (EVs) are very limited. Multi-Access Edge Computing (MEC) has been proposed to process the tasks generated by vehicles, which can reduce the latency and save the battery energy of EVs. However, the computing resource of MEC servers is still limited and cannot meet the delay requirements if massive EVs all offload the tasks. In addition, due to the uneven spatial and temporal distribution of vehicle arrivals, some MEC servers are busy, while some others are idle, resulting in the low resource efficiency and task completion ratio. In this paper, we propose the mobility-aware task offloading strategy method to allocate the computation resource of roadside servers for multiple EVs. We formulate the mathematical model of task offloading and resource allocation to jointly optimize computation latency and EV energy. Finally, the discrete particle swarm optimization algorithm is used to solve the problem. Simulation results show that the proposed method significantly alleviates the energy consumption and reduce the latency compared with conventional methods.

The development of software-defined industrial networks (SDIN) promotes the programmability and customizability of the industrial networks and is suitable to cope with the challenges brought by new manufacturing modes. For building more scalable and reliable SDIN, a distributed control plane with multicontroller collaboration becomes a promising option. However, as the brain of SDIN, the security of the distributed control plane is rarely considered. In addition to suffering direct attacks, each controller is also subjected to attacks propagated by other controllers because of information sharing or management domain takeover, resulting in the spread of attacks in a wider range than a single controller. Therefore, in this article, we study attacks against SDIN with distributed control plane, demonstrate their propagation across multiple controllers, and analyze their impacts. To the best of our knowledge, we are the first to study the security of SDIN with distributed control plane. In addition, since the existing defense mechanisms are not specifically designed for distributed SDIN and cannot defend it perfectly, we propose an attack mitigation scheme based on deep reinforcement learning to adaptively prevent the spread of attacks. Specifically, the novelty of our scheme lies in its ability of learning from the environment and flexibly adjusting the switch takeover decisions to isolate the attack source, so as to tolerate attacks and enhance the resilience of SDIN.

The data privacy and confidentiality in Internet-of-Things (IoT) networks have been one of the most concerned problems due to increasing threats. The commonly utilized IoT chips adopt a fixed authentication and encryption scheme in the link layer even though multiple options are usually supported. As different authentication and encryption operations mean dissimilar protections and various energy consumption, the fixed security strategy neglects the remaining energy, dynamic threats, and diverse service requirements, leading to low energy efficiency. Moreover, fixed high-level security protections consume too much energy even though the security requirement may be low, which results in a short working time. To address this problem, we propose an artificial intelligence (AI)-based adaptive security specification method for 6G IoT networks where the IoT devices are connected to cellular networks via different frequency bands, including terahertz (THz) and millimeter wave (mmWave). The IoT sensing devices are assumed to support the energy harvesting technique which is expected to be widely adopted in 6G. In our proposal, the extended Kalman filtering (EKF) method is first adopted to predict future harvesting power. Then, in each energy-aware cycle, we design a mathematical model to calculate the required energy of different security strategies and choose the supported highest level protection which can meet service requirement and avoid energy exhaustion. The simulation results illustrate that the proposal can not only provide satisfied security protection for different services but also adjust the security protection to avoid the energy exhaustion, leading to a significant improvement of throughput and working time.

Deep learning has emerged as a popular Artificial Intelligence (AI) technique to make conventional cyber physical systems become intelligent and sustainable. Recently, deep learning has been widely used in the network domain. With the aid of powerful deep neural networks, the communication network can carry out packets forwarding actions intelligently to avoid possible failure and congestion. However, with the high computing cost and process limitation in only the static network scenario, the existing deep learning based network traffic control algorithms cannot satisfy the sustainable requirement of next generation large scale dynamic network. To conquer the existing problems, a novel spatial-temporal value network aided deep learning based intelligent traffic control algorithm referred as ST-DeLTA is proposed in this paper. In ST-DeLTA, the value matrix and spatial temporal training model (ST model) are employed to intelligently extract the spatial as well as temporal features of traffic patterns and make adaptive packets forwarding decision in large scale and dynamic networks. The mathematical analysis gives the computing cost reduction of our proposal, and the computer simulation demonstrates that our proposal has significantly better training and network performance compared with traditional algorithms in terms of training accuracy, transmission throughput, and average packets loss rate.

Disasters, such as earthquakes, typhoons, and tsunamis, usually cause extreme damages to the communication infrastructures, which results in a heavy recovery workload and seriously affects people's life. The disaster recovery networks play a critical role to reduce the loss caused by the disasters. However, the suddenly varying traffic demand and limited resources after disasters may lead to the repetitive reconfigurations for running the existing packet forwarding strategies, such as the shortest path algorithms. To handle this problem, it is necessary to adopt the deep learning technique to develop a disaster-resilient solution. In this paper, we utilize the deep reinforcement learning technique to propose a self-adaptive routing method for the Movable and Deployable Resource Unit (MDRU) based backbone network. Compared with existing deep learning based routing strategy, our proposal can adapt to the sudden network errors. Moreover, we also analyze the deployment manner and consider a centralized control structure to significantly balance the traffic.

The Internet of Things (IoT) technology has attracted widespread attention since it can connect a huge number of heterogeneous devices to construct an intelligent environment to offer various services. Driven by the decreasing expense and significant convenience, it is expected that the total number of IoT devices will reach as high as 50 billion by 2020. And, these devices will provide diversified Internet services, including smart home, elder care, and vehicle-to-everything (V2X) communications. For the IoT technology, the network performance as well as the energy efficiency are both important since most end devices are battery limited. In this article, we focus on the IoT network based on the long-range wide-area network (LoRaWAN) protocol since its chirp modulation technology can adopt different spreading factors (SFs) to realize flexible communications. We propose a novel resource allocation strategy which utilizes multilayer virtual cell-based spatial time division multiple access (STDMA) scheme. This new method can not only simplify the calculation of the interference but also enables the cell radius to be adjusted to fit the communication distance. Moreover, the relationship between the power consumption and the network data rate is also analyzed in this article. We numerically analyze the optimal power consumption to achieve the highest data rate. The final performance evaluation demonstrates the significant improvement of our proposal compared with the conventional method.

Recently, the rapid advancement of high computing platforms has accelerated the development and applications of artificial intelligence techniques. Deep learning, which has been regarded as the next paradigm to revolutionize users' experiences, has attracted networking researchers' interests to relieve the burden due to the exponentially growing traffic and increasing complexities. Various intelligent packet transmission strategies have been proposed to tackle different network problems. However, most of the existing research just focuses on the network related improvements and neglects the analysis about the computation consumptions. In this paper, we propose a Value Iteration Architecture based Deep Learning (VIADL) method to conduct routing design to address the limitations of existing deep learning based routing algorithms in dynamic networks. Besides the network performance analysis, we also study the complexity of our proposal as well as the resource consumptions in different deployment manners. Moreover, we adopt the Heterogeneous Computing Platform (HCP) to conduct the training and running of the proposed VIADL since the theoretical analysis demonstrates the significant reduction of the time complexity with the multiple GPUs in HCPs. Furthermore, simulation results demonstrate that compared with the existing deep learning based method, our proposal can guarantee more stable network performance when network topology changes.

The fast increasing chip processing capacities driven by the Moore's Law have encouraged the academia and industry to consider more about general hardware architectures since they allow the repeated use for multiple purposes through the installations of applications. Some techniques utilizing the general hardware architectures have been developed to improve the flexibility of computer networks, such as the Software Defined Networking (SDN) and the Network Functions Virtualization (NFV). For these networks, the applications are required to be computation/communication-efficient since the installed applications share the hardware. In this paper, we study the resource-limited disaster recovery networks constructed by the Movable and Deployable Resource Units (MDRUs) which consist of various general computation platforms. We propose an efficient safety confirmation method through the photo sharing by the survivors. In the proposal, the Absorbing Markov Chain is utilized to model the safety confirmation process, transition matrix of which can be adopted to choose the suitable photo size for optimizing the traffic overhead and buffer consumption. Through periodical update of the photo database, unnecessary packet transmissions can be further avoided with reasonable sacrifice of the computation overhead. To expedite the computation, the GPU-accelerated MDRU is considered to conduct the matrix calculations in a parallel fashion.

The high-speed transmission has become extremely important with the rapid growth of network traffic in wireless networks. Because the available bandwidth of wireless channels are limited, Partially Overlapping Channels (POCs) are widely used in wireless networks to maximize the utilization of channel resources. However, with the traffic patterns of wireless networks becoming huge and dynamic, conventional POC assignment algorithms only designed for constantly generated network traffic are not suitable for the new generation wireless networks. Therefore, in this article, a joint deep Covolutional Neural Network (CNN) and activity vector based intelligent channel assignment algorithm is proposed, which is referred to as CNNAV. With the proposed CNNV approach, the network can learn from the historical traffic patterns and intelligently assign POCs to wireless links. The simulation result shows that, the network performance of our proposal in terms of both packets loss rate and network throughput are better than conventional POC assignment algorithms.

Due to the fast increase of sensing data and quick response requirement in the Internet of Things (IoT) delivery network, the high speed transmission has emerged as an important issue. Assigning suitable channels in the wireless IoT delivery network is a basic guarantee of high speed transmission. However, the high dynamics of traffic load (TL) make the conventional fixed channel assignment algorithm ineffective. Recently, the software defined networking-based IoT (SDN-IoT) is proposed to improve the transmission quality. Besides this, the intelligent technique of deep learning is widely researched in high computational SDN. Hence, we first propose a novel deep learning-based TL prediction algorithm to forecast future TL and congestion in network. Then, a deep learning-based partially channel assignment algorithm is proposed to intelligently allocate channels to each link in the SDN-IoT network. Finally, we consider a deep learning-based prediction and partially overlapping channel assignment to propose a novel intelligent channel assignment algorithm, which can intelligently avoid potential congestion and quickly assign suitable channels in SDN-IoT. The simulation result demonstrates that our proposal significantly outperforms conventional channel assignment algorithms.

Recently, network operators are confronting the challenge of exploding traffic and more complex network environments due to the increasing number of access terminals having various requirements for delay and package loss rate. However, traditional routing methods based on the maximum or minimum single metric value aim at improving the network quality of only one aspect, which makes them become incapable to deal with the increasingly complicated network traffic. Considering the improvement of deep learning techniques in recent years, in this paper, we propose a smart packet routing strategy with Tensor-based Deep Belief Architectures (TDBAs) that considers multiple parameters of network traffic. For better modeling the data in TDBAs, we use the tensors to represent the units in every layer as well as the weights and biases. The proposed TDBAs can be trained to predict the whole paths for every edge router. Simulation results demonstrate that our proposal outperforms the conventional Open Shortest Path First (OSPF) protocol in terms of overall packet loss rate and average delay per hop.

Recent years, Software Defined Routers (SDRs) (programmable routers) have emerged as a viable solution to provide a cost-effective packet processing platform with easy extensibility and programmability. Multi-core platforms significantly promote SDRs' parallel computing capacities, enabling them to adopt artificial intelligent techniques, i.e., deep learning, to manage routing paths. In this paper, we explore new opportunities in packet processing with deep learning to inexpensively shift the computing needs from rule-based route computation to deep learning based route estimation for high-throughput packet processing. Even though deep learning techniques have been extensively exploited in various computing areas, researchers have, to date, not been able to effectively utilize deep learning based route computation for high-speed core networks. We envision a supervised deep learning system to construct the routing tables and show how the proposed method can be integrated with programmable routers using both Central Processing Units (CPUs) and Graphics Processing Units (GPUs). We demonstrate how our uniquely characterized input and output traffic patterns can enhance the route computation of the deep learning based SDRs through both analysis and extensive computer simulations. In particular, the simulation results demonstrate that our proposal outperforms the benchmark method in terms of delay, throughput, and signaling overhead.