Objective Vehicle-infrastructure cooperative Networks(V2X)are open and contain large numbers of nodes with high mobility,frequent topology changes,unstable wireless channels,and varied service requirements.These characteristics create challenges to efficient data transmission.A flexible network that supports rapid reconfiguration to meet different service requirements is considered essential in Intelligent Transportation Systems(ITS).With the development of programmable network technologies,programmable data-plane techniques are shifting the architecture from rigid designs to adaptive and flexible systems.In this work,a protocol standard based on geospatial information is proposed and combined with a polymorphic network architecture to design a geospatial identifier network modal.In this modal,the traditional three-layer protocol structure is replaced by packet forwarding based on geospatial identifiers.Packets carry geographic location information,and forwarding is executed directly according to this information.Addressing and routing based on geospatial information are more efficient and convenient than traditional IP-based approaches.A vehicle-infrastructure cooperative traffic system based on geospatial identifiers is further designed for intelligent transportation scenarios.This system supports direct geographic forwarding for road safety message dissemination and traffic information exchange.It enhances safety and improves route-planning efficiency within V2X. Methods The geospatial identifier network modal is built on a protocol standard that uses geographic location information and a flexible polymorphic network architecture.In this design,the traditional IP addressing mechanism in the three-layer network is replaced by a geospatial identifier protocol,and addressing and routing are executed on programmable polymorphic network elements.To support end-to-end transmission,a protocol stack for the geospatial identifier network modal is constructed,enabling unified transmission across different network modals.A dynamic geographic routing mechanism is further developed to meet the transmission requirements of the GEO modal.This mechanism functions in a multimodal network controller and uses the relatively stable coverage of roadside base stations to form a two-level mapping:"geographic region-base station/geographic coordinates-terminal."This mapping supports precise path matching for GEO modal packets and enables flexible,centrally controlled geographic forwarding.To verify the feasibility of the geospatial identifier network modal,a vehicle-infrastructure cooperative intelligent transportation system supporting geospatial identifier addressing is developed.The system is designed to facilitate efficient dissemination of road safety and traffic information.The functional requirements of the system are analyzed,and the business processing flow and overall architecture are designed.Key hardware and software modules are also developed,including the geospatial representation data-plane code,traffic control center services,roadside base stations,and in-vehicle terminals,and their implementation logic is presented. Results and Discussions System evaluation is carried out from four aspects:evaluation environment,operational effectiveness,theoretical analysis,and performance testing.A prototype intelligent transportation system is deployed,as shown in Figure 7 and Figure 8.The prototype demonstrates correct message transmission based on the geospatial identifier modal.A typical vehicle-to-vehicle communication case is used to assess forwarding efficiency,where an onboard terminal(T3)sends a road-condition alert(M)to another terminal(T2).Sequence-based analysis is applied to compare forwarding performance between the GEO modal and a traditional IP protocol.Theoretical analysis indicates that the GEO modal provides higher forwarding efficiency,as shown in Fig.9.Additional performance tests are conducted by adjusting the number of terminals(Fig.10),background traffic(Fig.11),and the traffic of the control center(Fig.12)to observe the transmission behavior of geospatial identifier packets.The results show that the intelligent transportation system maintains stable and efficient transmission performance under varying network conditions.System evaluation confirms its suitability for typical vehicle-infrastructure cooperative communication scenarios,supporting massive connectivity and elastic traffic loads. Conclusions By integrating a flexible polymorphic network architecture with a protocol standard based on geographic information,a geospatial identifier network modal is developed and implemented.The modal enables direct packet forwarding based on geospatial location.A prototype vehicle-infrastructure cooperative intelligent transportation system using geospatial identifier addressing is also designed for intelligent transportation scenarios.The system supports applications such as road-safety alerts and traffic information broadcasting,improves vehicle safety,and enhances route-planning efficiency.Experimental evaluation shows that the system maintains stable and efficient performance under typical traffic conditions,including massive connectivity,fluctuating background traffic,and elastic service loads.With the continued development of vehicular networking technologies,the proposed system is expected to support broader intelligent transportation applications and contribute to safer and more efficient mobility systems.

In this paper, we introduce a novel straggler detection method based on the communication traffic model in the LLM training network, with the aim of identifying the straggler during the large language model (LLM) training. The theoretical communication traffic models for the data parallelism and pipeline parallelism are firstly constructed by analyzing communication characteristics to establish a expectation behavior of normal parallelism. Then, by comparing the captured traffic size with the theoretical model, the abnormal GPU pairs could be determined by using statistic threshold and dynamic variance analysis so as to avoid false alarm due to transient fluctuations. Furthermore, a GPU state matrix is set up to localize stragglers based on anomaly ratios. Simulation results demonstrate that the proposed communication traffic is consistent with the actual communication behavior and the proposed detection method could accurately identify straggler in short time.

Concurrency control (CC) scheme based on transaction decomposition has significantly enhanced the concurrency performance of multicore in-memory databases, surpassing traditional CC schemes such as two-phase locking (2PL) or optimistic concurrency control (OCC), particularly in high-contention scenarios. However, this performance improvement introduces new challenges, as balancing transaction dependency constraints with enhanced concurrency optimization remains a persistent issue, especially with the increased number of concurrent client requests, which can lead to complex transaction dependencies. To address these challenges, we propose Dynamic Contention Scheduling (DCoS), a novel method that enhances transaction concurrency via a dual-granularity architecture. DCoS integrates a deep reinforcement learning (DRL)-based executor to schedule high-contention transactions while preserving dependency correctness. DCoS employs a one-shot execution model that enables fine-grained scheduling in high-contention scenarios, while retaining lightweight in-partition execution under low-contention conditions. The experimental results on both micro- and macro-benchmarks demonstrate that DCoS achieves a throughput up to three times higher than state-of-the-art CC protocols under high-contention workloads.

Network security situation (NSS) prediction has attracted significant attention in recent years due to its potential to preemptively mitigate various types of network attacks. However, existing methods still suffer from several drawbacks, including slow convergence, susceptibility to local optima, and limited generalization ability, particularly when dealing with non-stationary and non-linear NSS data. In this paper, we propose a novel iterative optimized RBF-NN method for NSS prediction. Our proposed method leverages a resource allocation network (RAN) to dynamically determine the optimal number of neurons in the hidden layer, ensuring a balance between model complexity and prediction accuracy. Moreover, we introduce a cross-model method with a genetic algorithm to compute the optimal weights for the RBF-NN model. Specifically, we come up with a chaos search strategy during the iterative optimization process to prevent the RBF-NN model from falling into a local extreme point. Experimental results demonstrate that our method achieves a significant improvement in prediction accuracy, with an increase of up to 86.6%, while reducing the training time by up to 29.2% compared to existing techniques. These improvements are achieved within a tolerable training time, making our method both efficient and effective for real-world NSS prediction tasks.

Health monitoring of optical modules in large-scale AI computing clusters faces critical challenges: the extreme scarcity of negative samples hindering model training, and coarse binary ("normal/faulty") classification preventing proactive maintenance by obscuring gradual degradation. To overcome these limitations, this paper presents a fine-grained health assessment framework with Transformer auto-encoders. It incorporates expert-guided anomaly detection, leveraging expert knowledge to annotate initial faults and training the auto-encoder solely on normal data, eliminating reliance on scarce fault samples. The Transformer captures long-term temporal dependencies in multivariate time-series, enabling a 6-level health system (L0-L5) based on reconstructed errors, covering from inactive to irreversible fault for proactive maintenance. To mitigate misjudgment risks, a two-level verification mechanism calibrates samples initially classified into high-severity levels (L4-L5) using physical parameter thresholds. Experiments on 16,335,000 samples (0.00015% faults) validate its effectiveness in identifying degradation, supporting proactive maintenance.

Remote Procedure Calls (RPC) load balancing is crucial to maximize server resource utilization and reduce tail latency in data center. Existing load balancing methods deploy a frontend server acting as a load balancer, responsible for requesting backend server workload and routing RPC requests to backend server. However, existing methods suffer from two issues. Firstly, they introduce additional latency to RPC by requiring the frontend server to request the resource utilization of backend servers. Secondly, as the number of RPC requests grows, the frontend server becomes a bottleneck in the system, limiting the scalability of the load balancing approach. Programmable switches have become an effective way to serve high traffic rates involving many concurrent flows. In this paper, we propose a novel RPC load balancing approach that enables the programmable switch to take over the RPC request load balancing function. By maintaining a local estimation of the distribution of the backend server workload, the switch can directly guide the selection of a suitable backend server. We formulate optimizing server selection policy as a Bandits with Knapsack (BwK) problem. Our objective is to strike a balance between maximizing the cumulative utilization of server resources and minimizing cumulative latency. Extensive experiments in diverse scenarios demonstrate that the proposed method can effectively improve server selection policy and boost RPC load balancing performance.

The rise of intelligent manufacturing has induced many new demands on network technology.The polymorphic network separates the network technology from the support environment, supporting the coexistence as well as flexible loading and operation of various network technologies in the same network support environment, and can provide diversified, customizable, and highly reliable network service capabilities.The network environment requirements of intelligent manufacturing were studied, and smart factories were taken as an example to design efficient technical solutions based on the polymorphic network for various application scenarios.Different network technologies were efficiently adapted to application scenarios such as collaborative research and development design, production line control, machine vision quality inspection, unmanned intelligent inspection, accident emergency response, etc.The study verified the technical advantages of applying polymorphic networks to intelligent manufacturing, which could provide strong support for promoting the transformation and upgrading of the manufacturing industry.

Recent Internet of Things (IoT) security incidents indicate that current IoT defense methods are insufficient to defend against DDoS attacks due to a lack of timely and plausible mitigation mechanisms. These methods may make wrong decisions on traffic throttling when network traffic patterns dynamically change. To swiftly and properly defend against DDoS attacks in IoT, we propose a novel deep reinforcement learning (DRL) based DDoS defense approach. Our goals are to both safeguard normal traffic and discard malicious traffic. To achieve these goals, we establish a software defined networking (SDN) based IoT network by replacing traditional IoT gateways at certain locations with SDN switches. The proposed method consists of network monitoring, reward evaluation, and defense policy execution. The DRL agent is able to monitor the whole network by utilizing various traffic features from the SDN-enabled IoT network framework. We propose an adaptive punishment based reward function to accelerate the learning procedure. We also propose an available bandwidth allocation algorithm to refine defense policy so as to protect more normal traffic. The experimental results demonstrate that the proposed method could determine optimal defense policy faster by around 200 episodes and forward more normal traffic by around 22%.

A continuous-time value iteration based learning method is proposed for constrained-input nonlinear nonzero-sum game in this paper. Most existing studies were based on policy iteration, and thus they require an initial admissible control policy as the initial condition or some proper control policy to make the states satisfy the persistent excitation (PE) condition. However, no mater the initial admissible control policy nor a PE satisfied control policy, they can not be derived by a general feasible way. Such difficulty of choosing control policy may limit the actual application. The proposed method is developed based on value iteration and the requirement of choosing proper control policy can be avoided. Moreover, since the control signal should always be designed within limits in practice, the constrained-input property is taken into consideration. Simulation results are displayed to show the effectiveness.

The router throttling mechanism provides us a chance to prevent DDoS attack proactively through rate-limiting suspicious traffic before effective detection mechanism. The existing search-based and learning-based studies are highly customized to server load and can hardly cope with the constantly changing server load and unseen scenarios. To address the problem above, we design a self-evolutionary DDoS defense system, DeepThrottle, based on deep reinforcement learning (DRL) and router throttling mechanism in software defined network (SDN). The experimental results demonstrate that the DeepThrottle improves the passing ratio of normal traffic to the victim server, and reduces the server load under unseen attack scenarios compared with the state-of-the-art RL-based method.

构建一种实时可靠的传输网络以解决现有工业网络中低时延、高可靠的大规模数据传输问题是工业互联网建设的难点.当前工业以太网大多针对特定应用开发,种类繁杂,无法实现大范围、跨类别设备统一地互联互通,而传统互联网的尽力而为无法实现实时稳定的通信性能.本文利用时间敏感网络理论,提出并实现了基于802.1Qbv队列优先级的流量整形策略与基于帧抢占的调度策略,开发一种基于IEEE 802.1的时间敏感网络交换机,评估了基于IEEE 802.1的队列调度机制与技术.实验结果表明,该队列调度机制能有效保障流量的实时传输,且传输时延可控,从而保障了流量的可靠通信.

With the recent increase in the number of Internet of things(IoT)services,an intelligent schedul-ing strategy is needed to manage these services.In this paper,the problem of automatic choreography of mi-croservices in IoT is explored.A type of reinforcement learning(RL)algorithm called TD3 is used to generate the optimal choreography policy under the framework of a softwaredefined network.The optimal policy is gradu-ally reached during the learning procedure to achieve the goal,despite the dynamic characteristics of the network environment.The simulation results show that compared with other methods,the TD3 algorithm converges faster after a certain number of iterations,and it performs bet-ter than other non-RL algorithms by obtaining the highest reward.The TD3 algorithm can effciently adjust the traffic transmission path and provide qualified IoT services.

Botnets provide a fundamental infrastructure for various kinds of network attacks, such as DDoS attack, etc. Detecting DDoS attack in botnet is a long-standing challenge. In this paper, we propose a novel method to detect DDoS attack in botnet through the analysis of packet forwarding and network traffic. The primary idea is to collect features from both overall network traffic and packet to describe the attack pattern and conduct a detection model with high accuracy. Firstly, apart from overall network traffic, we calculate several statistics related to looking up functions of flow tables during packet forwarding on switches to describe attack pattern. Secondly, these features are put into a deep learning model to detect DDoS attack. We perform evaluations under Software Defined Networking (SDN) paradigm. In particular, we compare the proposed method with traditional methods. The experimental results demonstrate that the proposed method is meaningful in improving the accuracy of DDoS attack detection by adding packet-level features extracted from packet forwarding.

Task scheduling is one of the most fundamental technologies in a cloud computing environment. The essence of resource scheduling is to determine the appropriate executor and provide the on-demand resources to meet the task overhead. Due to the heterogeneity of the Internet of things (IoT), it has been a challenging task to efficiently schedule various IoT application tasks. An efficient scheduling system not only has to comply with task dependency constraints of IoT applications, but also should ensure the load balance of heterogeneous processors and minimize the total task execution time (i.e. makespan). In this paper, a collaborative management and scheduling framework is proposed based on an Adaptive Priority-based Particle Swarm Optimization (AP-PSO) for processing and communication resource allocation in heterogeneous IoT systems. The proposed approach is well adapted to the characteristics of IoT, including the processor heterogeneity, non-negligible communication costs, the diverse demand for service resources, and time-varying states of resources. Makespans and costs are adopted as metrics to evaluate the proposed scheduling model. Experimental results demonstrate that our approach outperforms other counterpart methods at the aspects of makespans and costs.

An adaptive dynamic programming value iteration algorithm is designed to solve nonlinear continuous-time nonzero-sum games in this paper. Since existing studies were developed on policy iteration, the initial condition of control policy requires to be admissible and thus the application has been limited. The proposed algorithm can be started with a more relaxed condition compared with policy iteration based methods, which relaxes the initial condition of the existing works. A simulation example is displayed to show the effectiveness.

Current Industrial Internet faces serious threats where attackers propagate malicious flows, resulting in communication failures in the Industrial Internet. In this work, we propose a practical and novel method to detect malicious traffic attack in real time with high accuracy. Our primary idea is to capture network flow, extract adequate network flow features, construct a long short-term memory (LSTM) based deep learning model, and identify the property of the corresponding network flow. Whether the network suffers attack or not is then determined according to the detection results. The corresponding prototype is also implemented in the Industrial Internet which is equipped with Software Defined Networking (SDN). Experimental results validate that the proposed method is effective in defending against malicious traffic attack in real-world network.

Cyberspace Mimicry Defense (CMD) has been widely used to achieve intrusion prevention against unknown system vulnerabilities or backdoors. The multi-ruling arbiter is a key part in CMD. This paper focuses on the problem of multi-ruling arbiter under data injection attack from the perspective of attacker and defender. We build a decentralized multi-ruling arbiter model for arbitration and introduced a standard iteration process to achieve consensus without attackers. We describe two data injection attack models for decentralized multi-ruling arbiter, namely random data injection attack and stealthy data injection attack. Further, we characterize the negative effect of the data injection attack on the performance of multi-ruling correctness. In order to mitigate the negative effect of random data injection attack, we propose a reliable multi-ruling arbitration approach based on adaptive threshold. By cutting future communication with the malicious neighbor, the decentralized multi-ruling arbiter is robust against random data injection attacks. Simulation results show that the proposed arbitration approach can effectively defend against random data injection attacks.

针对现有的拟态控制器防御成本过高的问题,研究了能够抵抗未知攻击且防御成本相对较低的拟态控制器.首先,定义信誉度与相异度表征执行体集的脆弱程度与异构程度,并从三大模块进行创新.执行体选择模块基于相异度与信誉度选择脆弱程度最低且异构程度最大的执行体集.然后,利用信誉度指标优化现有的多模裁决.最后,通过引入负反馈模块动态更新各执行体信誉度,实现执行体集选择策略与多模裁决策略的动态更新.

In this paper, we investigate the problem of distributed target tracking in wireless sensor networks (WSNs) under Byzantine data attacks. The dynamics of the target is defined by an evolution process. An M-ary quantizer is used at sensors to obtain local measurement data. With collected local measurement data from sensors in the network, the fusion center (FC) implements the target tracking process by using unscented kalman filter (UKF). For Byzantine nodes, the attack manner is described by the quantization process, the cascade of a normal mapping function and an attack function. We analyze the effect of Byzantine data attack on the performance of the distributed target tracking in terms of posterior Cramer-Rao lower bound (PCRLB). By making the FC obtain no information from both target state evolution model and reported data, we derive the condition to make the FC incapable of estimating the target location correctly and propose the corresponding strategy for the attacker. Numerical results show that the derived condition and proposed strategy can invalidate the distributed target tracking.

The distributed localisation with M-ary quantised data at sensor nodes is investigated in the presence of Byzantines from the attacker's perspective. The Byzantine nodes attack the network by transmitting falsified data to the fusion centre (FC) to deteriorate the performance of the distributed target localisation, where the independent and probabilistic Byzantine attack model is considered. By making the determinant of Fisher information matrix be zero, the optimal Byzantine attack strategy, which makes the FC incapable of estimating target location using the data at sensor nodes, is found. Numerical results show that the security performance of the distributed localisation improves sharply as the number of quantisation bits increases. Moreover, the security performance also increases tremendously as the attack probability decreases. Hence, the Byzantine attacker faces a trade-off between the performance damage and the hazard to be identified.