Proceedings of the 2nd International Conference on Networks, Communications and Intelligent Computing (NCIC 2024)

Table of Contents

1. Networks

   1. Frontmatter

   2. Capsule Network Based on Attention Mechanism for 6G Wireless Service Classification

      Xinyi Wang, Yuexia Zhang, Yang Hong, Shaoshuai Fan

      Abstract

      Accurate service classification can enable the efficient utilization of 6G wireless resources to improve system efficiency. This paper proposes a capsule network based on attention mechanism (CNBAM) for 6G wireless service classification to address the problem of traditional classification networks having difficulty achieving high-accuracy and fine-grained classification of 6G wireless services. CNBAM introduces the CBAM attention mechanism, which weighs the refinement of traffic flow features in spatial and channel dimensions to improve their differentiation. The unique EM-Routing algorithm of the capsule network is used to effectively prevent feature fizziness caused by pooling and improve the classification accuracy of the classification network. The simulation results show that CNBAM has high accuracy and stable classification for all classes.

   3. Enhanced Synchronization in Vehicular Digital Twin Networks: A Constrained Deep Reinforcement Learning Algorithm

      Hanji Wang, Xiaoshi Song, Pan Li, Ruiheng Zhang, Zhengbin Jiao

      Abstract

      With the progress of autonomous vehicles (AVs), the vehicular digital twin has emerged as an effective solution for enhancing the safety and reliability of AVs. One fundamental challenge in vehicular digital twin networks lies in the effective data synchronization between physical vehicles (PVs) and digital twins (DTs) under resource constraints. To tackle this challenge, this paper proposes a novel synchronous paradigm incorporating sampling, communication, and prediction components to enhance synchronization performance under transmission cost constraints. Particularly, we employ a cutting-edge concept, age of information, as the performance evaluation criterion and optimize synchronization performance by formulating a resource-constrained average AoI minimization problem. We transform it into a constrained Markov decision process and introduce a constrained deep reinforcement learning-based solution, namely the sampling, communication, and prediction co-design algorithm. Simulation outcomes reveal that our approach delivers enhanced synchronization performance with lower transmission costs than the baseline. As far as we know, this is the first paper to establish a unified framework for the co-design of sampling, communication, and prediction in vehicular digital twin networks.

   4. Nearest Neighbor Based Flow-Data Augmentation and Group Voting for Darknet Traffic Detection

      Chengpeng Dai, Jie Luo, Lulin Ni, Junyuan Zhang, Qingbing Ji

      Abstract

      With the rapid advancement of network technologies, darknet traffic has become a common channel for various illegal activities. Identifying darknet traffic and understanding the underlying user behaviors have become as critical challenges in the field of cybersecurity. Existing methods for darknet traffic detection largely focus on individual flows, neglecting the potential interdependencies between different flows. In this paper, we propose a novel approach for darknet traffic detection that combines nearest-neighbor flow data augmentation with a group voting mechanism, applicable to any flow-based darknet detection method. In the data processing phase, we utilize the nearest-neighbor algorithm to aggregate similar flow features, creating new samples and addressing the class imbalance inherent. Additionally, in the test phase, we introduce a weighted group voting mechanism based on neighborhood samples to predict the traffic type. This mechanism accounts for the relationships between neighboring flows, reducing the uncertainty of individual predictions and improving both model robustness and generalization performance. Experimental results demonstrate that the proposed method excels across multiple darknet datasets, achieving an average improvement of 2.3% in accuracy and 2.7% in F1-score.

   5. AoI-Sensitive Data Collection in UAV-Assisted Wireless Powered Communication Networks

      Zhaoyuan Wang, Zheng Zhou, Juan Liu

      Abstract

      This paper studies the age of information (AoI) minimization problem in an unmanned aerial vehicle (UAV)-assisted wireless powered communication network. Multiple time-constrained UAVs fly to positions within the transmission coverages of the sensor nodes (SNs) to charge them via wireless power transfer (WPT) in the downlink, and collect data from the SNs in the uplink when enough energy have been harvested. Given the dense deployment of SNs within the network, the transmission coverages of different SNs might overlap with each other. We are thus inspired to select some overlapped areas as hovering points (HPs), at which the UAVs conduct WPT and data collection with multiple SNs simultaneously. Accordingly, the HP selection, the wireless resource allocation and the trajectory planning for each UAV should be jointly considered in the AoI minimization problem. Since the optimization problem is nonconvex and thus difficult to be solved, a heuristic three-step algorithm is proposed. Firstly, a joint graph theory and kernel K-means algorithm is proposed to determine the number and positions of the HPs, and put them into different clusters, where each group of HPs is visited by a single UAV. Secondly, convex optimization methods are utilized to solve the wireless resource allocation subproblem. Finally, a time-constrained trajectory planning algorithm is proposed. Simulation results demonstrate that the introduced approach outperforms the traditional methods through effective utilization of overlapped areas.

   6. Performance Analysis of LEO Multi-satellite Cooperative Transmission Based on Stochastic Geometry

      Yaohua Sun, Ruiwen Li, Zixu Song, Ruifang Li, Shijie Zhang, Mugen Peng

      Abstract

      To enhance coverage and improve transmission rate, multi-satellite cooperative transmission is promising. However, analysis related to its performance is still lacking. Therefore, we consider a multi-satellite joint transmission scenario where the typical user can be served by multiple satellites, and spherical poisson point process is adopted to model low-earth-orbit (LEO) satellites. Then, we derive the expression of coverage probability by using stochastic geometry. The correctness of theoretical derivations is confirmed and the influence of network design parameters is illustrated by extensive simulation. Moreover, the performance improvement brought by multi-satellite cooperation is also highlighted.

   7. The Joint Coverage Probability of Vehicular Networks Based on Transdimensional Poisson Point Process

      Fengning Yang, Xiaoshi Song, Huimin Wei, Pan Li, Xinxin Yu

      Abstract

      This paper proposes an analytical framework based on stochastic geometry theory for investigating the joint coverage probability of two locations (\(\ell _1\) and \(\ell _2\)) in a vehicular networks. Particularly, we utilize the Transdimensional Poisson Point Process (TPPP) to model the actual distribution of base stations (BSs), where TPPP is a combination of 1D PPP and 2D PPP. Then, assume that the typical vehicle moves between the two locations \(\ell _1\) and \(\ell _2\) at a distance of v. With typical vehicle follows the closest BS association policy, two scenarios can occur: (i) the typical vehicle is associated with a new closer BS while moving to \(\ell _2\), (ii) when the previous BS is still the nearest after moving, the vehicle is served by the same BS in both \(\ell _1\) and \(\ell _2\). Our research shows that under the closest BS association policy, the joint coverage probability decreases with the increase of v, and the joint coverage probability is only the product of the individual coverage probabilities when v approaches infinity. In addition, we also investigate the effects of a series of network and channel parameters (e.g., BS density and \({\text {SIR}}\) target) on the joint coverage probability. These results provide important theoretical basis and practical reference for understanding coverage performance in vehicular networks.

   8. Topological Optimization and Routing for Reliable Communication Networks: A New Heuristic Method

      Jiabao Chen, Jiachen Sun, Jinyi Chen, Shuiqiang Li, Chao Fang, Zhuwei Wang, Chunyu Pan, Yuexia Zhang, Siyu Zhang

      Abstract

      To strengthen the network communication reliability influenced by critical node failure, a heuristic-based network topology optimization scheme is proposed in this paper. Through optimizing the location of some critical nodes, and using the shortest path routing method, we establish the maximum connectivity rate model to describe the network topology optimization problem improving the network reliability. In an optimization stage of critical nodes, considering the poor risk resistance of the neighboring nodes, and the low communication efficiency of a small degree node, a network topology optimization algorithm based on link adjustment and link addition coupled is proposed, which couples the link adjustment and link addition to perform the optimal position of critical nodes with low burden. Experimental results show that the proposed algorithm can improve 10% the network connectivity rate with low delay in a 14\(\times \)14 chessboard network. Furthermore, the proposed algorithm can be used to solve the critical issues in software-defined networks (SDN) to improve the network reliability and efficiency.

   9. Complex Network Attack Detection Strategy Based on Graph Neural Networks

      Jiangang Lu, Yaoxin Pan, Xiaozhi Deng, Tianming Huang

      Abstract

      As network attack techniques become increasingly sophisticated and stealthy, traditional detection methods are struggling to cope with emerging threats. This paper proposes a complex network attack detection strategy based on a multi-layer Graph Neural Network (GNN), which incorporates adaptive graph structure modeling and multi-scale feature aggregation techniques. By dynamically adjusting the network topology and capturing multi-level attack features, the proposed strategy enables precise identification and classification of complex attack behaviors. Extensive experiments on multiple public datasets demonstrate that this approach significantly outperforms traditional detection techniques, particularly in handling advanced persistent threats (APT) and other sophisticated attack scenarios, exhibiting strong adaptability and robustness. The findings suggest that this detection strategy not only expands the theoretical application of GNNs in network security but also provides an innovative technical solution to addressing complex network attacks in practice .

   10. Multi-tier-leadership Stackelberg Games for Caching Equilibrium in Vehicle Networks

       Ruyu Xu, Yufang Zhang, Mianmian Dong, Chunze Jia, Peng Wang, Chen Chen, Ling Xu

       Abstract

       Overlooking the dynamic nature of vehicle paths and the suddenness of vehicular network services, traditional popularity-based caching strategies tend to concentrate data on specific base stations (BSs), leading to imbalanced cache distribution. To address this problem, a balanced caching mechanism for heterogeneous vehicular networks based on multi-tier-leadership Stackelberg games is proposed. In the game process, based on hierarchical relationship of network architecture, BSs act as leaders, road side units (RSUs) as access points subordinate to the BS as sub-leaders, and vehicles as followers. By integrating the utility functions of leaders, sub-leaders, and followers in multi-tier-leadership Stackelberg games framework to derive optimal pricing and caching strategies. Simulation results demonstrate that the total system profit converges after approximately 200 iterations. Additionally, this mechanism improves cache stability by 16.2% and reduces transmission delay by 38% compared to the DQN algorithm.

   11. Large-scale Agile Earth Observation Satellite Scheduling Based on Window Decision Network

       Lin Ma, Kangning Du, Benkui Zhang, Jun Kang, Peiran Song

       Abstract

       The large-scale agile earth observation satellite scheduling problem is a complex combinatorial optimization problem. Existing approaches typically address this by adjusting the sequence of observation missions, which can lead to a limited solution space and restricted optimization potential. To overcome these limitations, we propose a Window Decision Network. It enhances the accuracy of the problem state representation by integrating both the static and dynamic attributes of the observation process. Additionally, we developed a visible time window update algorithm to guide the construction of the solution sequence, expanding the solution space by adjusting the window sequence directly. Experimental results demonstrate that this method outperforms the state-of-the-art approaches, increasing scheduling profit by up to 23.163%.

   12. Research on DDoS Attack Detection Model Based on Transformer Architecture

       Boyi Liang

       Abstract

       In recent years, the era of large-scale models has dawned upon our daily lives, with the majority of smartphone vendors integrating these sophisticated technologies into their devices’ systems. Smartphones have become an inseparable part of our existence, housing private assets, information, and intimate details of our lives. Consequently, the threat of attacks targeting personal mobile devices has escalated, rendering cybersecurity a paramount concern of our times. To address this pressing issue, this paper presents a deep neural network, leveraging the ubiquitous Transformer architecture, specifically tailored to identify network intrusions. The proposed model, through its intricate design and powerful capabilities, has achieved a remarkable accuracy of 98.54% when evaluated on the esteemed NF-UQ-NIDS-V2 dataset. This achievement underscores the potential of Transformer-based solutions in fortifying the cybersecurity landscape, safeguarding users’ data and privacy against the ever-evolving landscape of cyber threats.

   13. Research on Bearing Defect Recognition Based on YOLOv5-CBAM

       Yuhang Qiu, Xinru Wei, Xiaoyu Jia, Xueke Fan, Xinsheng Xu, Na Wang, Shan Wang

       Abstract

       As industrial automation and intelligence advance progressively, the necessity for enhanced precision in bearing defect identification and detection is on the rise. Deep learning intelligent algorithms facilitate the convenient and sophisticated recognition of bearing defects, addressing current demands effectively. However, prevalent models are encumbered by substantial parameter counts and suffer from suboptimal accuracy, compromising the reliability of defect identification processes. This study investigates the problem of bearing defect detection by employing a YOLOv5-CBAM-based methodology for recognition purposes. The integration of the CBAM (Convolutional Block Attention Module) attention mechanism into the architecture of the YOLOv5s backbone network has facilitated the development of an enhanced YOLOv5 model, specifically optimized for the purpose of bearing defect detection. The analysis of experimental outcomes indicates that the enhanced algorithm introduced in this study achieves a significant reduction in model complexity, with the total number of parameters reduced to 6,709,997. This represents a decrease of 316,310 parameters in comparison to the YOLOv5s algorithm model. The mean Average Precision at 50% (mAP50) achieved an accuracy of 0.791, representing a 0.6% enhancement relative to the accuracy of the original model .

   14. Passive Indoor Localization Based on a Hybrid Model Using Deep Residual and Temporal Convolutional Networks

       Yunling Chen, Aohao Wang, Yujia Han, Tuanwei Tian, Hao Deng, Dongming Zhang, Rui Ma

       Abstract

       This work investigates the issue of using channel state information (CSI) collected from a single link to achieve passive indoor personnel positioning in intricate indoor environments. Frequency features and temporal features are extracted separately for the input data using a hybrid model of a deep residual network (ResNet) and temporal convolutional network (TCN). First, we adaptively filter the amplitude information to remove outliers and perform phase calibration on the phase information to recover the true phase. Then, we employ a deep ResNet to extract frequency features in CSI data and enhance the focus on global features by adding a convolutional block attention module (CBAM). Furthermore, we utilize a TCN to extract temporal features in CSI data, to capture local and long-term dependencies efficiently. Finally, we experimentally verify the feasibility of this method using a CSI acquisition platform built with TP-LINK routers and desktop computers equipped with Intel 5300 network cards.

   15. Networking Design and Performance Evaluation for ISAC in 6G Cellular Networks

       Lincong Han, Yahui Xue, Jing Dong, Jing Jin, Qixing Wang

       Abstract

       Integrated sensing and communication (ISAC) is a new feature for future sixth generation (6G) cellular networks. There are four sensing modes, including monostatic, multiple monostatic, bistatic and multistatic, the networking schemes of which is a key issue in the ISAC networks. In this paper, we propose two types of networking schemes for them, in consideration of both interference elimination and sensing performance. One is the “fish scale” networking scheme and the other is “ring shape”. The former select one cell from each site with the same beam direction as the sensing transceiver for the first two modes to avoid both the inter-site and the inter-cell interference. The latter set all the three cells within one site as sensing transmitters (Txs) or receivers (Rxs) for the last two modes, and adjacent sites play different roles when sensing. We also design area division for targets at different positions of the network. The system-level simulations show the signal-to-noise ratio as well as the sensing accuracy for different networking schemes, and verifies the performance gain by involving more sensing Txs and Rxs.

   16. A Learnable Vulnerability Mining Model for Internet of Things

       Pengbin Hu, Lingling Tan, Zhen Zhang

       Abstract

       With the rapid development of information technology, the widespread use of IoT devices has increasingly highlighted security issues. Existing vulnerability discovery techniques often focus on optimizing single processes, lacking systematic organization and application of discovery experiences. Therefore, this paper proposes a learnable vulnerability discovery framework inspired by genetic principles in biology. It encodes key feature data from HTTP interactions into DNA and builds a knowledge tree model with the discovery experiences, integrating multiple discovery experiences to form a knowledge forest. By calculating the similarity between the target and the DNA chains in the knowledge forest, the framework guides the vulnerability discovery process, enabling effective inheritance and utilization of experiences, thus endowing the process with memory and experience transfer capabilities. The DNAFuzzer model is used to validate this approach. Results show that when discovering vulnerabilities in different device models with code reuse, DNAFuzzer can precisely locate and trigger vulnerabilities within one minute based on past discovery experiences. Its average discovery efficiency improves by 83.79% compared to boofuzz, 92.65% compared to sulley, and 94.43% compared to peach.

   17. Joint Task Offloading and Resource Allocation Strategy for MEC Enabled LEO Satellite Networks

       Yuchen Cai, Pei Gao, Xiankui Luo, Chao Cai, Zhaoyang Su, Xianglong Duan, Liu Liu

       Abstract

       By introducing Mobile Edge Computing (MEC) into satellite networks enables the sinking of cloud computing power closer to users and improves user experience. However, the limited resources of edge servers cannot meet the huge demand for computing resources in remote areas, which may lead to the increase in task processing delay and reduce the communication efficiency. Therefore, efficient resource allocation strategies and task offloading decisions are needed to reduce the total delay of on Orbit Edge Computing systems. Based on the above problems, this paper proposes a joint resource allocation and task offloading strategy, which is mainly divided into two parts, the optimal resource allocation through Lagrange multiplier method, and the optimization of offloading strategy for different service types through the improved Grey Wolf Optimization. Simulation results demonstrate that the proposed algorithm can reduce the total system delay effectively.

   18. Chip Atomic Clock Placement Strategy for Low-Earth Orbit Satellite Time Synchronization

       Hongzhou Zhang, Jiaen Zhou, Yafei Zhao

       Abstract

       Low-Earth Orbit (LEO) satellite systems have rapidly advanced in various fields, including space target detection, mobile communication, and navigation. High-precision time references are essential for accurate navigation and positioning, with chip atomic clocks offering a viable solution for establishing high-precision time-frequency references for LEO satellites. This paper proposes a K-Medoids clustering-based strategy for optimal deployment of chip atomic clocks, considering synchronization performance, frequency stability, and cost. Simulation results validate the strategy’s effectiveness, showing significant improvements in clock offset accuracy and frequency stability compared to high-stability crystal oscillators, random placement, and K-Means clustering, reducing the system clock offset to 10.153 ns, with improvements of 55.30%, 36.25%, and 18.82%, respectively.

   19. Energy-Efficient Design of UAV-Assisted Hierarchical Federated Learning

       Jie Zhang, Juan Liu, Zheng Zhou

       Abstract

       In recent years, a distributed machine learning framework called federated learning (FL) has received much attention. However, in order to ensure the accuracy of the training model, the user device needs to perform multiple rounds of local computation and constantly communicate with the central server to update the model, thereby consuming a significant amount of energy. To reduce energy consumption at user devices, this paper proposes a hierarchical federated learning (HFL) framework assisted by unmanned aerial vehicles (UAVs), where UAVs are employed as edge aggregators to receive and aggregate the models from user devices and relay the updated model to the data center for global aggregation after certain rounds of local training and edge aggregation. The UAV-assisted HFL problem is constructed as a nonlinear mixed integer programming (MIP) problem that aims to minimize the user device energy consumption by jointly optimizing the user-UAV association, the UAVs’ positions, and the transmit power at each user device. Subsequently, an iterative algorithm based on block coordinate descent (BCD) is proposed to find the optimal solution. Simulation results demonstrate that the proposed UAV-assisted HFL method can significantly reduce the total energy consumption at user devices while ensuring the training accuracy.

   20. A Framework for Blockchain Network Topology Reduction Based on Network Protocol Parsing

       Zhixin Meng, Yang Li, Gang Xiong, Zhen Li, Gaopeng Gou, Zhong Guan

       Abstract

       Blockchain network topology reduction can assist cybersecurity researchers in understanding network structures and establishing protective mechanisms. However, existing topology reduction methods face two key challenges: limited attention to routing relationship information within the network and inaccurate node detection, making effective validation difficult. In the paper, we propose a measurement framework for blockchain network topology reduction. The framework leverages network protocol analysis to enhance the effectiveness of topology reduction. Through the three modules of node active discovery, node online verification and attribute resolution, and node route detection, we detect the node information and routing relationship information in the blockchain network, so as to perform topology reduction of the blockchain base network. We conducted experiments on two typical blockchain networks, Bitcoin and Ethereum, and the results demonstrate that our method is more efficient and comprehensive than comparison methods within the same time frame.

   21. Energy and Distance Weighted Dynamic Clustering Protocol for Wireless Sensor Networks

       Ruiyan Han, Jinglun Shi, Farhad Banoori

       Abstract

       Wireless sensor networks (WSNs) usually encounter significant obstacles in extending the network longevity due to energy scarcity and large-scale random deployment of sensors. Hierarchical cluster-based routing protocols like LEACH that can prolong network survival are well-suited for WSNs. However, LEACH adopts a probabilistic function to randomly select the cluster head (CH), most likely resulting in the choice of nodes with low energy levels being picked as CHs and an uneven distribution of CHs. To address the randomness in CH selection, we propose the energy and distance weighted dynamic clustering (EDDC) protocol for WSNs, which combines control parameters that encompass residual energy, distance from nodes to the base station (BS), and the percentage of optimal number of clusters in the CH selection mechanism. Consequently, EDDC gives priority to the node possessing the greatest residual energy that is geographically close to BS as the CH to equalize energy expenditure. Simulation results indicate that our proposed EDDC surpasses LEACH protocol when it comes to residual energy and network lifespan, showing its effectiveness in enhancing the endurance of WSNs.

   22. A Lightweight Authentication and Key Agreement Method for Satellite Communication Based on Identity-Based Cryptography

       Guoyi Zhang, Kai Wang, Chong Wang, Xiaoyang Liang, Changqing Lai, Yun Bai, Hongyan Xu, Hao Qi, Ying Yang, Xingxing Wang, Xiang Zhu

       Abstract

       Satellite communication constitutes an essential part of the modern communication domain. With its extensive application, security issues in satellite communication, such as spoofing attacks, data theft, data tampering, and replay attacks, have become increasingly prominent. As a crucial defense in network security protection, identity authentication can effectively identify the identity of each node in the network, achieve two-way authentication and security key negotiation between satellites and users, and resist illegal attacks. This paper presents a lightweight authentication method for satellite communication based on SM9, the Chinese National Cryptography Standard for Identity-Based Cryptography. This approach takes SM9 as the core and combines the characteristics of the satellite communication system to design a comprehensive set of identity authentication and key agreement methods. The experimental results demonstrate that the proposed method meets the requirements of low latency and high security in the satellite communication environment, providing a strong guarantee for the security of satellite communication.

   23. The Study of Distributed UDSF Architecture for AMF Edge Deployment in NPN

       Lin Yuan, Zhan Xu, Zhigang Tian

       Abstract

       The demand for low latency and high privacy in 6G applications has driven the deployment of AMF closer to the edge, enabling distributed deployments to achieve faster local response, meet the privacy and security requirements of Non-Public Networks (NPN), and enhance network reliability. However, as AMF moves closer to the edge, the frequency of user context transfer between AMF instances significantly increases, leading to higher system overhead, slower response times, and challenges in maintaining data consistency. To address these challenges, this paper proposes a storage architecture based on distributed UDSF. Distributed UDSF distributes user context data across multiple nodes in the network, allowing AMF instances to access data sources nearby, effectively reducing cross-instance data transfer, lowering transmission latency, and ensuring user data security and privacy, thereby reducing latency and improving fault tolerance while achieving system stability and scalability .

   24. The Red Ocean Heuristic Algorithm for Large-Scale Agile Earth Observation Satellite Scheduling

       Pengfei Gao, Lin Cao, Benkui Zhang, Yu Liu

       Abstract

       The Agile Earth Observation Satellite Scheduling Problem is a complex NP-hard challenge commonly tackled using heuristic algorithms. Traditional heuristic algorithms often exhibit low task completion rates and inefficiency, particularly in large-scale scenarios. This paper presents the red ocean heuristic algorithm, which improves both completion rates and computational efficiency through iterative decision-making, modeled on trader behavior. Experimental results show that our algorithm outperforms traditional heuristic, achieving a 12.13% improvement in completion rate and a 66.93% reduction in computation time.

   25. A Heuristic Algorithm for Mobility-Aware Task Offloading in Vehicular Edge Computing

       Wei Dong, Fan Jiang, Junxuan Wang, Xuewei Zhang

       Abstract

       Vehicular Edge Computing (VEC) can deliver low-latency and high-reliability services to vehicle users. In order to address the challenge of frequent changes in vehicle locations and to enhance the integration of MEC and IoV technologies, this paper presents a simulation of a dynamic VEC network, taking into account various task attributes, vehicle mobility patterns, and time delay constraints. The optimization objective aims to identify edge servers that satisfy the delay constraints imposed by the vehicles’ mobility trajectories while minimizing energy consumption during the task offloading procedure. To achieve this goal, we propose a heuristic mobility-aware offloading algorithm (HMAOA), which continuously updates the resource selection and offloading policies to adapt to changes in vehicle locations. Simulation results demonstrate that the proposed algorithm effectively reduces task processing energy consumption, enhances the task completion rate, and adapts well to dynamic environment.

   26. Performance Trade-Off and Analysis of Integrated Sensing and Communication in Vehicular Networks

       Xinxin Yu, Xiaoshi Song, Huimin Wei, Pan Li, Fengning Yang

       Abstract

       Integrated Sensing and Communication (ISAC) is a critical technology in vehicular networks, which combines the sensing and communication capabilities in one system. Due to limited resources such as spectrum and transmit power, one essential problem lies in the study of ISAC-based vehicular networks is the performance trade-off between sensing and communication. In this paper, by applying tools from stochastic geometry, we first model the vehicular network as the Cox process driven by a Poisson line process (PLP). Then, we derive the detection probability of the sensing module and the coverage probability of the communication module, respectively. We further capture the respective performance trade-off between sensing and communication under the bandwidth budget based on the derived analytical results. Finally, extensive simulations are conducted to verify the accuracy of our derivations. This research is essential for optimizing resource allocation and improving service quality within ISAC in vehicular networks. By focusing on the integration of sensing and communication capabilities, we aim to advance the development of smarter vehicular networks that can better meet the needs of users.

   27. Dehazing Using GAN Network Based on YUV Color Space

       Zhe Zhang, Chang Lin, WenXing Zou, Bing Fang

       Abstract

       Although image dehazing algorithms based on a single Generative Adversarial Network (GAN) effectively remove haze, they still encounter issues such as dark brightness and blurred details. This paper proposes a dual-generator dehazing structure based on a GAN network, integrating channel and spatial attention mechanisms. By generating dehazed and brightness-restored images in both RGB and YUV color spaces and adjusting the contrast between the two sets of images, the brightness and details are effectively restored, resulting in final output images that are more natural and consistent. Experimental results demonstrate that this method offers significant improvements over existing dehazing algorithms on multiple public datasets, exhibiting superior performance in both subjective visual effects and objective metrics.