Started in January,1974(Monthly)
Supervised and Sponsored by Chongqing Southwest Information Co., Ltd.
ISSN 1002-137X
CN 50-1075/TP
CODEN JKIEBK
Editors

Featured ArticlesMore...

  • Volume 51 Issue 3, 15 March 2024
      
      Information Security Protection in New Computing Mode
      Overview of IoT Traffic Attack Detection Technology Based on Fuzzy Logic
      SHANG Yuling, LI Peng, ZHU Feng, WANG Ruchuan
      Computer Science. 2024, 51 (3): 3-13.  doi:10.11896/jsjkx.230700130
      Abstract ( 51 )   PDF(2244KB) ( 75 )   
      References | Related Articles | Metrics
      The Internet of things(IoT) is progressively permeating our daily activities,interconnecting an array of diverse physical devices to the Internet.This foundational connectivity underpins applications spanning smart cities,e-health,precision agriculture,and beyond.The swift proliferation of IoT applications,however,has been paralleled by an upsurge in the frequency of network attacks targeting these devices and services.The complex and dynamic nature of these attacks,coupled with their imprecision and uncertainty,has significantly compounded the intricacies of accurate detection and identification.In response to these exigencies,a novel approach has emerged in the form of fuzzy logic-based attack detection frameworks.These frameworks strategically integrate varied fuzzy techniques throughout diverse operational phases to facilitate heightened precision in the detection of network attacks,particularly in instances characterized by data inaccuracy and uncertainty.Within the expanse of this comprehensive survey paper,a meticulous exposition unfolds.It commences by delving deeply into the realm of IoT security,dissecting its multifaceted dimensions,such as the security challenges it responds to,the required security requirements,and the types of attacks it faces.Subsequently,it offers a detailed portrayal of intrusion detection systems(IDS) and further encapsulates the foundational framework of IDS within the IoT domain.The foundational tenets of fuzzy logic are subsequently expounded upon,followed by a discerning analysis of the rational underpinning the integration of fuzzy logic in traffic attack detection.In subsequent sections,a discerning comparative analysis of diverse traffic attack detection schemes,grounded in disparate technological methodologies,is meticulously presented.This analytical elucidation underscores their respective performance metrics and,by extension,their pivotal significance within this burgeoning sphere.Finally,the synthesis of the principal contributions encapsulated within this paper is meticulously articulated,concurrently outlining pathways for future research.These nascent trajectories are expected to provide researchers with new perspectives and enrich the academic discourse to mitigate escalating cyberattacks.
      Contrastive Graph Learning for Cross-document Misinformation Detection
      LIAO Jinzhi, ZHAO Hewei, LIAN Xiaotong, JI Wenliang, SHI Haiming, ZHAO Xiang
      Computer Science. 2024, 51 (3): 14-19.  doi:10.11896/jsjkx.230800063
      Abstract ( 43 )   PDF(2068KB) ( 76 )   
      References | Related Articles | Metrics
      Misinformation proliferates on the Internet,undermining the normal functioning of various industries.Detecting falsehoods accurately has therefore become an urgent challenge.Existing research on this task focuses primarily on three aspects:account traits,textual content,and multimodality.However,most methods overlook the key attribute of misinformation diffusion the novelty of content.They analyze the veracity of target claims in isolation,failing to capture public opinion dynamics.To address this issue,this paper proposes a cross-document misinformation detection framework called contrastive graph learning(CAL).CAL focuses on content novelty and comprises two key components:a contrastive learning module and a heterogeneous graph module.The former expands the representational difference between factual and false claims,and the latter encompasses five entity types:words,events,event sets,sentences,and documents.It injects semantic features of the public discourse into entity embeddings.We evaluate CAL on the IED,TL17,and Crisis datasets at both document and event levels.CGL achieves state-of-the-art performance,which verifies the efficacy of its design.It provides a robust solution for combating misinformation by mode-ling novelty and environmental context.
      Survey of Incentive Mechanism for Cross-silo Federated Learning
      WANG Xin, HUANG Weikou, SUN Lingyun
      Computer Science. 2024, 51 (3): 20-29.  doi:10.11896/jsjkx.230700194
      Abstract ( 38 )   PDF(1876KB) ( 50 )   
      References | Related Articles | Metrics
      As a kind of distributed machine learning,federated learning effectively solves the problem of data sharing in big data era.Among them,cross-silo federated learning,as a type of federated learning in which institutions cooperate with each other,is obviously very important to design a reasonable incentive mechanism in the process of cross-silo cooperation.Based on the perspective of cross-silo cooperation,this paper makes a comprehensive analysis of the existing incentive mechanism of cross-silo fe-derated learning.Firstly,this paper introduces three basic problems in the process of cross-silo cooperation:high privacy,data he-terogeneity,and fairness.Then,it analyzes the incentive mechanism design methods under two different cross-silo cooperation models centered on the global model and centered on participants.Finally,it summarizes the several factors that affect the stable development of cross-silo cooperation:data evolution of participants,changes in the cooperative relationship of participants,and negative behaviors of participants,and looks forward to the future direction of cross-silo federal cooperation.
      Differential Privacy Data Synthesis Method Based on Latent Diffusion Model
      GE Yinchi, ZHANG Hui, SUN Haohang
      Computer Science. 2024, 51 (3): 30-38.  doi:10.11896/jsjkx.230700177
      Abstract ( 36 )   PDF(3004KB) ( 46 )   
      References | Related Articles | Metrics
      The widespread application of data sharing and publication in the socio-economic domain drives scientific progress and societal development.However,issues related to copyright and privacy,especially concerning personal data,remain critical challenges.Differential privacy data synthesis has emerged as an effective means of protecting data privacy,where data holders can release synthetic data instead of real data,thereby enhancing data utility and availability while preserving privacy.In response to the limited usability of existing differential privacy generation models,this paper proposes a two-stage differential privacy generation model based on the latent space diffusion approach.Firstly,the differential privacy-aware information compression is performed on the original image,and it is projected from the pixel space to the latent space to obtain the desensitized latent vector representation of the original sensitive data.The latent vector is then fed into a diffusion model to gradually transform into a prior distribution and sampled through a denoising process.Experimental results based on the MNIST and Fashion MNIST datasets demonstrate that the proposed model exhibits significant improvements in terms of Fréchet inception distance(FID) and downstream task accuracy compared to state-of-the-art models like DP-Sinkhorn.
      Study on Blockchain Based Federated Distillation Data Sharing Model
      LIU Wei, LIU Yuzhao, TANG Congke, WANG Yuanyuan, SHE Wei, TIAN Zhao
      Computer Science. 2024, 51 (3): 39-47.  doi:10.11896/jsjkx.230700186
      Abstract ( 23 )   PDF(3006KB) ( 39 )   
      References | Related Articles | Metrics
      The privacy of raw data makes it difficult to be directly shared among multiple participants.The issue of data security sharing and privacy-preserving has become a hot research topic.To solve this problem,this paper proposes a blockchain-based bederated distillation data sharing model(BFDS).It utilizes blockchain to form a collaborative teacher network with multiple participants.Through distilled output exchange,the knowledge from complex teacher networks is transferred and used to train lightweight models.A novel multi-weight node trust evaluation algorithm is proposed that uses smart contracts to generate traceable global soft labels.It can reduce the negative impact caused by quality differences among participants.Experimental results show that BFDS can collaborate with multiple parties to share data knowledge reliably,distill training models collaboratively,and reduce model deployment costs.The proposed algorithm can effectively reduce the negative impact of low-quality nodes and improve the quality and security of global soft labels.
      Database & Big Data & Data Science
      Framework and Algorithms for Accelerating Training of Semi-supervised Graph Neural Network Based on Heuristic Coarsening Algorithms
      CHEN Yufeng , HUANG Zengfeng
      Computer Science. 2024, 51 (3): 48-55.  doi:10.11896/jsjkx.221200158
      Abstract ( 22 )   PDF(1576KB) ( 33 )   
      References | Related Articles | Metrics
      Graph neural network is the mainstream tool of graph machine learning at the current stage,and it has broad development prospects.By constructing an abstract graph structure,the graph neural network model can be used to efficiently deal with problems in various application scenarios,including node prediction,link prediction,and graph classification.But the application on large-scale graphs has always been the key point and difficulty in graph neural network training.And how to effectively and quickly train and deploy graph neural networks on large-scale graph data is a major problem hindering the further industrial application of graph neural networks.Graph neural network can use the topological information of the network structure of the graph,so as to achieve better results than other general neural networks such as multi-layer perceptron on the node prediction problem.But the rapid growth of the number of nodes and edges of the graph's network structure restricts the training of the graph neural network,and the number of nodes in the real dataset is tens of millions or even billions,or the number of edges in some dense network structures has reached tens of millions.This makes it difficult for traditional graph neural network training methods to achieve direct results.This paper improves and proposes a new framework for graph neural network training based on heuristic graph coarsening algorithms,and proposes two specific training algorithms on this basis.Then this paper proposes two simple heuristic graph coarsening algorithms.Under the guarantee that the loss of accuracy is acceptable and the memory space consumption is greatly reduced,the proposed algorithm can further significantly reduce both calculation time and training time of graph neural networks.Experiment shows that satisfactory results can be achieved on common datasets.
      Large-scale Multi-objective Evolutionary Algorithm Based on Online Learning of Sparse Features
      GAO Mengqi, FENG Xiang, YU Huiqun, WANG Mengling
      Computer Science. 2024, 51 (3): 56-62.  doi:10.11896/jsjkx.230100004
      Abstract ( 25 )   PDF(2353KB) ( 27 )   
      References | Related Articles | Metrics
      Large-scale sparse multiobjective optimization problems(SMOPs) are widespread in the real world.Proposing generic solutions for large-scale SMOPs can improve problem-solving in the fields of evolutionary computation,cybernetics,and machine learning.Due to the high-dimensional decision space and the sparse Pareto-optimal solutions of SMOPs,existing evolutionary algorithms are vulnerable to the curse of dimensionality when solving SMOPs.To address these problems,a large-scale multi-objective evolutionary algorithm based on online learning of sparse features(MOEA/OLSF) is proposed,with the learning of sparse distribution as an entry point.Specifically,an online learning sparse features method is designed to mine nonzero variables.Then a sparse genetic operator is proposed for further searching nonzero variables and generating offspring solutions.Its binary crossover and mutation operators are used to control the sparsity and diversity of solutions in the nonzero variable mining process.The comparison results with the state-of-the-art algorithms on test problems with different scales show that the proposed algorithm outperforms the existing algorithm in terms of convergence speed and performance.
      Study on Task Analysis Methods Based on Attention-GAN
      ZHOU Linru, PENG Pengfei
      Computer Science. 2024, 51 (3): 63-71.  doi:10.11896/jsjkx.221100012
      Abstract ( 15 )   PDF(3338KB) ( 24 )   
      References | Related Articles | Metrics
      A reasonable task analysis can help decision makers to plan tasks quickly and accurately.The current task analysis method using case-based reasoning has problems such as long analysis time and low accuracy of analysis results.The method uses LSTM as the generator and RNN as the discriminator.For failing to return updates for small gradients of discrete data,the ge-nerator uses rollout policy to complete the incomplete sequence generated and the discriminator uses Monte Carlo(MC) to sample the data to obtain the complete data sequence action value function,thus guiding on updating the parameters of the generator.To address the problems of sparse data with obscure features and unclear data focus,a soft attention mechanism is added to the GAN before training.It assigns different weights to different features to filter out redundant data and select the important features.The proposed method is compared with the GAN without the attention mechanism on the same simulated dataset,and it is demonstrated that the method with the attention mechanism improves 0.088,0.092,0.094 and 0.068 in terms of P,R,F1 value and accuracy respectively.Compared with other neural network recommendation algorithms, P,R,F1 valueand accuracy is improvedby 0.1~0.3,0.1~0.2,0.1~0.25 and 0.07~0.17,respectively,which proves the effectiveness of the proposed method.
      Traffic Speed Forecasting Algorithm Based on Missing Data
      HUANG Kun, SUN Weiwei
      Computer Science. 2024, 51 (3): 72-80.  doi:10.11896/jsjkx.230100045
      Abstract ( 18 )   PDF(2498KB) ( 23 )   
      References | Related Articles | Metrics
      Traffic speed forecasting is the foundation of intelligent transportation system,which can ease traffic congestion,save public resources and improve people's quality of life.In real situations,the collected traffic speed data are usually missing,and most of the existing research results only consider the scenarios with relatively complete data.The paper focuses on the traffic speed data in the missing scenarios,captures the spatio-temporal correlation,and predicts the future traffic speed.In order to make full use of the spatio-temporal characteristics of traffic data,this study proposes a new deep learning-based traffic speed forecasting model.Firstly,a “recover-predict” algorithm is designed,which first uses a self-supervised learning method to enable the model to recover the missing data and then predict the traffic speed.Secondly,a contrastive learning method is introduced to make the feature representation of the speed time series more robust.Finally,the scenarios with different missing data rates are simulated,and experimental results show that the prediction accuracy of the proposed method outperform existing methods with various missing rates,and experiments are designed to analyze the comparative learning method and different recovery algorithms to prove the effectiveness of the proposed method.
      Academic Influence Ranking Algorithm Based on Topic Reputation and Dynamic HeterogeneousNetwork
      CHEN Pan, CHEN Hongmei, LUO Chuan
      Computer Science. 2024, 51 (3): 81-89.  doi:10.11896/jsjkx.230100037
      Abstract ( 13 )   PDF(2796KB) ( 26 )   
      References | Related Articles | Metrics
      Effectively mining academic big data and analyzing academic influence of papers are benefical for researchers to obtain important information.The dynamic changes of text content and academic network structure have an important impact on the ranking results of academic impact.However,the existing ranking algorithms of academic influence of papers either lack consideration of text contents or the dynamic changes of academic network structure.To solve this problem,this paper proposes an algorithm for ranking academic influence,which is called TND-Rank,based on topic reputation and dynamic heterogeneous network.In TND-Rank,the impact of the topic on the paper at a certain time is measured and embedded to the paper influence ranking algorithm that takes into account the time factor.The dynamic ranking related to the academic impact of a paper is calculated by comprehensively considering the influence of various factors,i.e,the level of topic prestige,journal,author,and time etc.In the experiments,the AMiner data set published between 1936 and 2014 with complete information are analyzed,and compared with four related algorithms in recent years.Spearman correlation coefficient,normalized discounted cumulative gain(NDCG) and graded average precision(GAP) are adopted to evaluate performance of the algorithm.Experimental results verify the feasibility and effectiveness of the proposed algorithm TND-Rank,which can effectively synthesize various information to rank the academic influence of papers.
      Community Search Based on Disentangled Graph Neural Network in Heterogeneous Information Networks
      CHEN Wei, ZHOU Lihua, WANG Yafeng, WANG Lizhen, CHEN Hongmei
      Computer Science. 2024, 51 (3): 90-101.  doi:10.11896/jsjkx.221200029
      Abstract ( 11 )   PDF(4234KB) ( 25 )   
      References | Related Articles | Metrics
      Searching the community containing a given query node in heterogeneous information networks(HINs) has a wide range of application values,such as friend recommendation,epidemic monitoring and so on.However,most of the existing HINs community search methods impose strict requirements on the topology of the community based on the predefined subgraph pattern,ignoring the attribute similarity between nodes,which will be difficult to locate the community with weak structural relationship and high attribute similarity.And the global search mode is difficult to effectively deal with large-scale network data.To solve these problems,we design disentangled graph neural network and the local modularity based on meta path to measure the attribute similarity and structural cohesion between nodes respectively.Moreover,we use the 0/1 knapsack problem to optimize the impact of the attribute and structure on the community,define the most valuable c-size community search problem,and then propose a value maximization community search algorithm based on disentangled graph neural network to perform a three-stage search process.In the first stage,we construct candidate subgraphs according to the query in-formation and meta-path,control the search range within the local range of the query vertex to ensure the search efficiency of the whole algorithm.In the second stage,we use the disentangled graph neural network to fuse the heterogeneous information and user label information to calculate the attribute similarity between nodes.In the third stage,we design a greedy algorithm to find the c-size community with high attribute similarity and structural cohesion according to the community definition and cohesion measurement indicator.Finally,we test the performance of algorithm on real homogeneous and heterogeneous data sets,and a large number of experimental results demonstrate the effectiveness and efficiency of the proposed model.
      Sequential Recommendation Based on Multi-space Attribute Information Fusion
      WANG Zihong, SHAO Yingxia, HE Jiyuan, LIU Jinbao
      Computer Science. 2024, 51 (3): 102-108.  doi:10.11896/jsjkx.230600078
      Abstract ( 14 )   PDF(2868KB) ( 25 )   
      References | Related Articles | Metrics
      The goal of sequential recommendation is to model users' dynamic interests from their historical behaviors,and hence to make recommendations related to the users' interests.Recently,attribute information has been demonstrated to improve the performance of sequential recommendation.Many efforts have been made to improve the performance of sequential recommendation based on attribute information fusion,and have achieved success,but there are still some deficiencies.First,they do not explicitly model user preferences for attribute information or only model one attribute information preference vector,which cannot fully express user preferences.Second,the fusion process of attribute information in existing works does not consider the in-fluence of user personalized information.Aiming at the above-mentioned deficiencies,this paper proposes sequential recommendation based on multi-space attribute information fusion(MAIF-SR),and proposes a multi-space attribute information fusion framework,fuse attribute information sequence in different attri-bute information spaces and model user preferences for different attribute information,fully expressing user preferences using multi-dimensional interests.A personalized attribute attention mechanism is designed to introduce user personalized information during the fusion process,enhance the personalized effect of the fusion information.Experimental results on two public data sets and one industrial private data set show that MAIF-SR is superior to other comparative sequential recommendation models based on attribute information fusion.
      Deep Collaborative Truth Discovery Based on Variational Multi-hop Graph Attention Encoder
      ZHANG Guohao, WANG Yi, ZHOU Xi, WANG Baoquan
      Computer Science. 2024, 51 (3): 109-117.  doi:10.11896/jsjkx.221200063
      Abstract ( 11 )   PDF(2850KB) ( 25 )   
      References | Related Articles | Metrics
      In the era of big data,the release of data value often requires the fusion of multi-source data,and data conflict has become an inevitable key problem in this process.In order to filter out true claims and reliable sources from conflicting data,researchers have proposed truth discovery methods.However,the existing truth discovery methods pay more attention to the direct collaborative information between sources and claims,and ignore the deeper indirect collaborative and confrontational information,which is insufficient to express the characteristics of sources and claims.To solve this problem,this paper proposes a truth discovery method based on variational multi-hop graph attention encoder(TD-VMGAE).It constructs a bipartite graph network based on the inclusion relationship between sources and claims,uses a multi-hop graph attention layer to gather indirect cooperative information and antagonistic information for of each node,and a truth discovery variational auto-encoder is designed to extract the categorical distribution required in node characterization,and collaborative classification of data sources and claims is carried out.Experiments show that the proposed method has good performance in three datasets with different scales,and the effectiveness and generalization ability of the method are verified by ablation experiments and visualization.
      Event Prediction Based on Dynamic Graph with Local Data Augmentation
      PAN Lei, LIU Xin, CHEN Junyi, CHENG Zhangtao, LIU Leyuan, ZHOU Fan
      Computer Science. 2024, 51 (3): 118-127.  doi:10.11896/jsjkx.221200054
      Abstract ( 20 )   PDF(2251KB) ( 18 )   
      References | Related Articles | Metrics
      Event refers to activities that occur in real world at specific time and places.For instance,unrest,violent terrorist attacks,natural disasters and the spread of infectious diseases,will bring great threats and losses to national security and human life.If the occurrence of such events could be predicted more precisely and effectively,the impact of negative events will be minimized,and it is possible to maximize the benefits of the positive events.It is still a very challenging task to predict events accurately.An event prediction method named local augmented temporal-GAT(LAT-GAT) based on graph attention network is proposed in this paper.It uses conditional variational encoders to generate new features,which will be concatenated with the original features to new one,based on neighbors of the current node.With this approach,our model can utilize the propagation structure of events.In addition,the chronological order of events occurrence is considered by our model.The feature of events in last time point is integrated into the output of the neural network in current time.The temporal property of event propagation is exploited through temporal data integration.And finally,the proposed method is compared with a number of representative baseline me-thods on the real-world datasets,including Thailand,India,Egypt and Russia.The results show that LAT-GAT has the best F1 scores in all datasets.The recall of our model exceeds that of any other baseline methods in the datasets of Thailand,Russia and India.In Thailand,Egypt and India,our model achieves the best precision.Ablation experiments are also conducted to investigate the influence of the model parameters on the final results.
      Time Series Completion and One-step Prediction Based on Two-channel Echo State Network
      ZHENG Weinan, YU Zhiyong, HUANG Fangwan
      Computer Science. 2024, 51 (3): 128-134.  doi:10.11896/jsjkx.221200055
      Abstract ( 15 )   PDF(1486KB) ( 27 )   
      References | Related Articles | Metrics
      With the development of the Internet of Things,numerous sensors can collect a large number of time series with rich data correlation,providing powerful data support for various data mining applications.However,some objective or subjective reasons(such as equipment failure,sparse sensing) often lead to the loss of collected data to varying degrees.Although many approaches have been proposed to solve this problem,data correlation is either not fully considered or computationally expensive.In addition,existing methods only focus on the completion of missing values,and fail to take into account downstream applications.Aiming at the above shortcomings,this paper designs a two-channel echo state network to achieve both the completion task and the prediction task.Although the two channels share the input layer,they have their own reservoir and output layer.The biggest difference between them is that the output layer of the left/right channels respectively represents the target value or prefilled va-lue corresponding to the moment before/after the input layer.Finally,by fusing the estimates of the two channels,the data correlation from before and after the missing moments is fully utilized to further improve performance.Experimental results of diffe-rent missing rates with two missing mechanisms(random missing and piecewise missing) show that the proposed model is superior to the current methods in both completion accuracy and prediction accuracy.
      Deep Neural Network Model for Transmission Line Defect Detection Based on Dual-branch Sequential Mixed Attention
      HAO Ran, WANG Hongjun, LI Tianrui
      Computer Science. 2024, 51 (3): 135-140.  doi:10.11896/jsjkx.230600109
      Abstract ( 16 )   PDF(2412KB) ( 23 )   
      References | Related Articles | Metrics
      Detecting defects in transmission lines and repairing them in a timely manner is of great practical significance for ensuring the safety and stability of the power grid.However,due to the complex background and small component size of transmission line images,existing object detection models cannot achieve satisfactory results.Therefore,this paper proposes a deep neural network model for detecting defects in transmission lines based on dual-branch serial attention.The model designs dual-branch serial attention(DBSA) to allow the model to focus more weight on the defects,and proposes well-connected feature pyramid network(WCFPN) to enable full fusion of the features extracted by DBSA,thereby improving the model's ability to detect small targets.DBSA compresses the feature map along the height and width branches and extracts attention using one-dimensional convolution to achieve fine-grained control over the features.WCFPN designs a new fusion path that includes cross-scale fusion and skip-layer connections,allowing high-level semantic information and low-level spatial information extracted by DBSA to interact more fully.Finally,experiments are conducted on five transmission line datasets,including insulator explosion,damaged anti-vibration hammer,bird's nest debris,broken cementpole and transmission line defect,and the proposed model achieves the best detection performance.The average AP50and AP of the five datasets is 84.3% and 46.1%,respectively,which is 3.7% and 3% higher than that of the state-of-the-art model YOLOv7.
      Computer Graphics & Multimedia
      Multi-view Autoencoder-based Functional Alignment of Multi-subject fMRI
      HUANG Shuo, SUN Liang, WANG Meiling, ZHANG Daoqiang
      Computer Science. 2024, 51 (3): 141-146.  doi:10.11896/jsjkx.230600166
      Abstract ( 13 )   PDF(1936KB) ( 21 )   
      References | Related Articles | Metrics
      One of the major challenges in functional magnetic resonance imaging(fMRI) research is the heterogeneity of fMRI data across different subjects.On the one hand,analyzing multi-subject data is crucial for determining the generalizability and effectiveness of the generated results across subjects.On the other hand,analyzing multi-subject fMRI data requires accurate anatomical and functional alignment among the neural activities of different subjects to enhance the performance of the final results.However,most existing functional alignment studies employ shallow models to handle the complex relationships among multiple subjects,severely limiting the modeling capacity for multi-subject information.To solve this problem,this paper proposes a multi-view auto-encoder functional alignment(MAFA) method based on multi-view auto-encoders.Specifically,our method learns node embedding by reconstructing the response spaces of different subjects,capturing shared feature representations among subjects,and creating a common response space.We also introduce the graph clustering process by introducing self-training clustering objectives using high-confidence nodes as soft labels.Experimental results on four datasets demonstrate that the proposed method achieves the best decoding accuracy compared to other multi-subject fMRI functional alignment methods.
      Unsupervised Low-light Image Enhancement Model with Adaptive Noise Suppression and Detail Preservation
      GAO Ren, HAO Shijie, GUO Yanrong
      Computer Science. 2024, 51 (3): 147-154.  doi:10.11896/jsjkx.221200074
      Abstract ( 12 )   PDF(4732KB) ( 31 )   
      References | Related Articles | Metrics
      The visual quality of images taken under low-light environment is usually low,due to many factors such as low lightness and imaging noise.Current low-light image enhancement methods have a common limitation that they only focus on improving lightness condition and suppressing noise,but neglect to preserve image details.To solve this problem,an unsupervised low light image enhancement method is proposed in this paper,aiming to improve the visibility and preserve the fidelity of an image with good efficiency.The model consists of two stages,i.e.,low-light enhancement and noise suppression.In the first stage,an unsupervised image decomposition module and a lightness enhancement module are constructed to achieve the goal of improving visibility.In the second stage,under the guidance of the illumination distribution of an image,we synthesize pairwise training data and train the denoising network to depress the imaging noise from the originally-dim regions and preserve the image details of the originally-bright regions.Compared with other methods,experimental results show that our method achieves better balance between the goals of visibility improvement and fidelity preservation.In addition,our method can be attractive in real-world applications,as it does not need to collect bright-dim image pairs,and it has small model size and fast calculation speed.
      Appearance Fusion Based Motion-aware Architecture for Moving Object Segmentation
      XU Bangwu, WU Qin, ZHOU Haojie
      Computer Science. 2024, 51 (3): 155-164.  doi:10.11896/jsjkx.221200153
      Abstract ( 15 )   PDF(4262KB) ( 27 )   
      References | Related Articles | Metrics
      Moving object segmentation aims to segment all moving objects in the current scene,and it is of critical significance for many computer vision applications.At present,many moving object segmentation methods use the motion information from 2D optical flow maps to segment moving objects,which have many defects.For moving objects moving in the epipolar plane or moving objects whose 3D motion direction are consistent with the background,it is difficult to identify these objects by the 2D optical flow maps.Besides,incorrect 2D optical flow also effects the result of moving object segmentation.To solve the above problems,this paper proposes different motion costs to improve the performance of moving object segmentation.In order to detect moving objects with coplanar and collinear motion,this paper proposes a balanced reprojection cost and a multi-angle optical flow contrast cost,which measures the difference between the 2D optical flow of moving objects and that of the background.For ego-motion degeneracy,this paper designs a differential homography cost.To segment moving objects in complex scenes,this paper proposes an appearance fusion based motion-aware architecture.In this architecture,in order to effectively fuse appearance features and motion features of objects,the multi-modality co-attention gate is adapted to achieve better interaction between appearance and motion cues.Besides,to emphasize moving objects,this paper introduces a multi-level motion based attention module to suppress redundant and misleading information.Extensive experiments are conducted on the KITTI dataset,the JNU-UISEE dataset,the KittiMoSeg dataset and the Davis-2016 dataset,and the proposed method achieves excellent performance.
      Object Detection Method with Multi-scale Feature Fusion for Remote Sensing Images
      ZHANG Yang, XIA Ying
      Computer Science. 2024, 51 (3): 165-173.  doi:10.11896/jsjkx.230200030
      Abstract ( 18 )   PDF(4906KB) ( 22 )   
      References | Related Articles | Metrics
      Object detection for remote sensing images is an important research direction in the field of computer vision,which is widely used in military and civil fields.The objects in remote sensing images have the characteristics of multiple scales,dense arrangement and similarity between classes,so that the object detection methods used in natural images have many omissions and false detection in remote sensing images.To address this problem,this paper proposes an object detection method with multi-scale feature fusion based on YOLOv5 for remote sensing images.Firstly,a residual unit fusing multi-head self-attention is introduced into the backbone network,through which multi-level feature information is fully extracted and semantic differences among diffe-rent scales were reduced.Secondly,a feature pyramid network fusing lightweight upsampling operators is introduced for obtaining high level semantic features and low-level detail ones.And the feature maps with richer feature information could be acquired by feature fusion,which improves the feature resolution of objects at different scales.The performance of the proposed method is evaluated on the datasets DOTA and NWPU VHR-10,and the accuracy(mAP) of the method isimproved by 1.5% and 2.0%,respectively,compared with the baseline model.
      Combined Road Segmentation and Contour Extraction for Remote Sensing Images Based on Cascaded U-Net
      LI Yu, YANG Xiangli, ZHANG Le, LIANG Yalin, GAO Xian, YANG Jianxi
      Computer Science. 2024, 51 (3): 174-182.  doi:10.11896/jsjkx.221200032
      Abstract ( 22 )   PDF(4205KB) ( 23 )   
      References | Related Articles | Metrics
      Aiming at the problem that the deep-learning-based model for road information extraction can only output single-task results and the inadequate use of correlation between multiple tasks,a combined road segmentation and contour extraction method based on cascaded U-Net is proposed,which extracts the road contour after fusing the feature map of road semantic segmentation with the original image.Firstly,the U-Net network structure is used to extract the hierarchical features of optical remote sensing images,and the cascaded U-Net structure is introduced to concatenate the features to extract the pixel-level label and contours of roads respectively.Secondly,the attention mechanism module is added to each stage of U-Net to extract spatial context information and deep level features to improve the detection sensitivity of details.Finally,the joint loss function composed of dice coefficient and cross-entropy error is used for the overall training to extract simultaneously the road semantic segmentation and contour results.On the optical remote sensing dataset of the urban area of Ottawa,Canada,the joint extraction method of road information based on cascaded U-Net achieves 42% precision,58% recall,48.2% F1 score and 71.6% mIoU in the segmentation index,and achieves a global optimal threshold(ODS) of 0.896 in the road detection index.The results show that,the model can meet the requirements of joint extraction of road multi-task information and has better detection accuracy.
      Artificial Intelligence
      Review of Reinforcement Learning and Evolutionary Computation Methods for StrategyExploration
      WANG Yao, LUO Junren, ZHOU Yanzhong, GU Xueqiang, ZHANG Wanpeng
      Computer Science. 2024, 51 (3): 183-197.  doi:10.11896/jsjkx.230400058
      Abstract ( 15 )   PDF(3628KB) ( 27 )   
      References | Related Articles | Metrics
      Reinforcement learning and evolutionary computation,as two types of nature-inspired learning paradigms,are the mainstream methods for solving strategy exploration problems,and the fusion of these two types of methods provides a general solution for solving strategy exploration problems.This paper analyzes the basic methods of reinforcement learning and evolutionary computation,the basic methods of strategy exploration,the fused methods of strategy exploration,and the frontier challenges in four aspects,and it is expected to bring inspiration to the cross-fertilization research in this field.
      Chinese Named Entity Recognition Based on Label Information Fusion and Multi-task Learning
      LIAO Meng, JIA Zhen, LI Tianrui
      Computer Science. 2024, 51 (3): 198-204.  doi:10.11896/jsjkx.230200114
      Abstract ( 22 )   PDF(1537KB) ( 26 )   
      References | Related Articles | Metrics
      With the development of Chinese named entity recognition research,most models focus on enriching feature representation by integrating vocabulary or glyph information but ignore label information.Therefore,a Chinese named entity recognition model integrating label information is proposed in this paper.Firstly,the embedding representation of characters is obtained by pre-trained model BERT-wwm,and labels are represented as vectors.The character representation and label representation are interactively learned by using the Transformer decoder structure to capture the interdependence between characters and labels and enrich the feature representation of characters.To promote the learning of label information,a supervision signal based on text sentences is constructed,multi-label text classification tasks are added,and multi-task learning is used for training.Among them,the named entity recognition task uses a conditional random field for decoding and prediction,and the multi-label text classification task uses a biaffine mechanism for decoding and prediction.The two tasks share all parameters except the decoding layer,which ensures that different supervision information is fed back to each subtask.Several groups of comparative experiments are carried out on the public data sets MSRA,Weibo,and Resume,and the F1 values of 95.75%,72.17%,and 96.23% are obtained respectively.Compared with several benchmark models,experimental result of the proposed model is improved to some extent,which validates its effectiveness and feasibility.
      Dual Feature Adaptive Fusion Network Based on Dependency Type Pruning for Aspect-basedSentiment Analysis
      ZHENG Cheng, SHI Jingwei, WEI Suhua, CHENG Jiaming
      Computer Science. 2024, 51 (3): 205-213.  doi:10.11896/jsjkx.230100035
      Abstract ( 15 )   PDF(2039KB) ( 25 )   
      References | Related Articles | Metrics
      Existing models use graph neural network based on dependency trees for aspect-based sentiment analysis,which improves the classification performanceof the model to a certain extent.However,due to technical limitations of dependency parsing,the inaccuracy of the dependency parsing results leads to a large amount of noise in the dependency tree,which makes the performance improvement of the model is limited.In addition,some sentences themselves do not conform to the standard syntactic structure.Previous studies utilized syntactic and semantic information with the same confidence level without fully considering the difference in their contributions to determining the polarity of aspect words,resulting in poor model performance on the corres-ponding datasets.To overcome these challenges,a dual feature adaptive fusion network based on dependency type pruning is proposed in this paper.Specifically,the model uses a novel hybrid approach,named dependency type pruning and adjacency matrix smoothing,to mitigate the noise generated by dependency parsing.In addition,the model fully considers the availability of syntactic information of sentences through a dual feature adaptive fusion module to combine syntactic features and semantic features for aspect-level sentiment analysis in a more flexible way.Extensive experiments on five publicly available datasets demonstrate that the proposed method significantly outperforms baseline models.
      Equilibrium Optimization Algorithm Based on Variable Generation Probability and Multi-difference Cauchy Variation
      LI Kewen, NIU Xiaonan, LI Guoqing, CUI Xueli
      Computer Science. 2024, 51 (3): 214-225.  doi:10.11896/jsjkx.221200129
      Abstract ( 13 )   PDF(3629KB) ( 17 )   
      References | Related Articles | Metrics
      In order to solve the problem that the standard equalization optimization algorithm(EO) lacks the balance ability of global search and local search and is easy to fall into local optimal,an equalization optimization algorithm(VDEO) based on variable generation probability and multi-difference Cauchy variation is proposed.First,the diversity of the initial population is increased with Tent chaotic mapping,which provides the basis for optimization.Secondly,the variable generation probability is introduced to replace the original fixed value,so that the algorithm can increase the global search ability in the early stage of iteration,and pay attention to the solving accuracy in the later stage,so as to improve the balance ability of global search and local search.Finally,the fusion of different difference strategies and Cauchy variation helps the optimization process to escape from the local optimal.Aiming at 15 benchmark test functions including single-peak,multi-peak and fixed-dimension multi-peak and CEC2022 test functions,VDEO and ten heuristic algorithms EO,GWO,WOA,SCA,MFO,AOA,AVOA,BWO,AHA and POA are simulated and compared under multiple dimensions.The Wilcoxon rank sum test is performed on the experimental results of the benchmark function.Experimental results show that VDEO achieves better global search and local search balance,and has better ability to jump out of the local optimal and higher convergence accuracy.
      Dynamic Path Planning Algorithm for Heterogeneous Groups in Aircraft Carrier Aviation SupportOperations
      SUN Didi, LI Chaochao
      Computer Science. 2024, 51 (3): 226-234.  doi:10.11896/jsjkx.221200119
      Abstract ( 10 )   PDF(3911KB) ( 23 )   
      References | Related Articles | Metrics
      The path planning task in aircraft carrier support operation has the problem of high dynamic scene and strong heterogeneity of agents.Although the traditional global path planning algorithm can obtain the global optimal results,it can not adapt to the highly dynamic changing scene,and can not solve the security problem caused by the heterogeneity of agents.The current local path planning algorithm can well solve the problem agent size difference,but it is difficult to express heterogeneous group behavior control uniformly.In order to solve the above problems,a dynamic path planning algorithm for heterogeneous groups in aircraft carrier aviation support operations is proposed.Firstly,the optimized global and local path planning algorithms are integrated to solve the highly dynamic problem of the aviation support operation scene.The path is adjusted in time according to the dynamic environment information,and the security problem caused by the highly dynamic scene to the heterogeneous agents are fully considered.Secondly,the method considers the different behavior characteristics of heterogeneous agents,and adopts the behavior control model of heterogeneous agents based on kinematics characteristics in the process of local collision avoidance.Finally,taking the American Nimitz as an example,and the algorithm is evaluated in aspects of path length,smoothness,security,obstacle avoidance ability by using UE4 simulation experiments.Simulation results show that,compared with other path planning algorithms,the proposed algorithm can not only generate safe paths for heterogeneous groups on aircraft carrier deck,but also meet the application requirements of heterogeneous groups in dynamic aviation support operation scenarios.
      TMGAT:Graph Attention Network with Type Matching Constraint
      SUN Shounan, WANG Jingbin, WU Renfei, YOU Changkai, KE Xifan, HUANG Hao
      Computer Science. 2024, 51 (3): 235-243.  doi:10.11896/jsjkx.221200097
      Abstract ( 15 )   PDF(2327KB) ( 21 )   
      References | Related Articles | Metrics
      The graph structure has recently been employed to solve the KGC problem of knowledge graph completion.The graph neural network(GNNs) constantly updates the representation of the central entity by aggregating the entity's local neighborhood information,whereas the graph attention network(GATs) focuses on aggregating the neighbors to obtain a more accurate representation of the central entity by using the attention mechanism.Although these models performed well in KGC,they all neglect the central entity's type information and only use neighborhood information to calculate attention,resulting in inaccurately mea-sured attention.This paper proposes a type-matching constrained graph attention network(TMGAT) to address these issues.The entity type-relation level of attention is derived by calculating the attention of the central entity type to each neighborhood relationship,and the type matching degree between the central entity and each neighborhood connection is further determined.The entity-relation level attention is then determined by combining the type matching degree through the neighborhood relation and the corresponding neighbor entity,and the final attention of the neighborhood node to the central entity is obtained.The type matching degree is utilized to constrain the traditional attention mechanism,increase attention mechanism accuracy,obtain more accurate central entity embedding,and subsequently improve knowledge graph completion accuracy.The proposed TMGAT is the first model of knowledge graph completion task combined with explicit type in GATs up to this point.To validate the TMGAT model's performance,two existing data sets are processed so that each entity in the data set has several types.Finally,experiment demonstrates TMGAT's high competitiveness in knowledge completion tasks and the effect of the number of types on the model's performance analyzed.
      Self-calibrating First Spike Temporal Encoding Neuron Model
      FENG Ren, CHEN Yunhua, XIONG Zhimin, CHEN Pinghua
      Computer Science. 2024, 51 (3): 244-250.  doi:10.11896/jsjkx.221200003
      Abstract ( 11 )   PDF(2046KB) ( 22 )   
      References | Related Articles | Metrics
      Because of the complex spatio-temporal dynamic process of spike neurons and the non-differentiable spike information,the training of spike neural network(SNN) has always been very difficult.The ANN-to-SNN method for indirect training of deep SNN avoids the difficulties of direct training of deep SNN.However,the performance of the SNN obtained in this approach is greatly affected by the spike information encoding mechanism.Among many coding mechanisms,TTFS has a good biological basis and is energy efficient,but existing TTFS codes use a single-spike formalism,which has weak information representation capability and large time windows for encoding.Therefore,based on the single spike coding of TTFS,a calibration spike is added to form a self-calibrating first spike time to first spike coding mechanism,and the corresponding SC-TTFS neuron model is constructed.In SC-TTFS,the first spike is the spike that must be emitted,while the calibration spike determines whether it is emitted according to the residual membrane potential after the first spike is emitted,which is used to compensate the quantification error and truncation error caused by the coding spike and to reduce the time window required for coding.The advantages of this approach are verified by comparing and analyzing the corresponding conversion errors of various codes and ANN-SNN conversion experiments on various network architectures.On CIFAR10 and CIFAR100 datasets,the proposed algorithm is verified by experiments based on VGG and ResNet network structures,and it achieves ANN-SNN transformation with non-destructive accuracy on both network structures and two data sets.Compared to state-of-the-art similar methods,the SNN constructed by the proposed method has the smallest network inference latency.In addition,on the VGG structure,the proposed method improves the energy efficiency by about 80% compared with TTFS coding.
      Knowledge Graph Embedding Model with Entity Description on Cement Manufacturing Domain
      ZHOU Honglin, SONG Huazhu, ZHANG Juan
      Computer Science. 2024, 51 (3): 251-256.  doi:10.11896/jsjkx.221200080
      Abstract ( 14 )   PDF(1904KB) ( 23 )   
      References | Related Articles | Metrics
      To address the problem that many knowledge graph embedding models lack the consideration of semantic information when performing knowledge embedding and cannot extract the semantic information of entities specialized in cement manufactu-ring domain well.The entity description text is added to the embedding work of cement manufacturing domain knowledge graph(CMFKG),and the knowledge graph embedding with entity description model(KGEED) is proposed,which adopts the TransE model to get the embedding of structural information of CMFKG.The CNN-based entity description embedding module is used to obtain the semantic-based embedding of CMFKG,and the triples of structural information embedding and semantic information embedding are fused with CNN,so that the rich entity description text information of the knowledge graph in cement manufactu-ring domain can be well considered.Experiments show that the model achieves good results in the embedding work of the know-ledge graph in cement manufacturing domain.
      Prediction of Lower Limb Joint Angle Based on VMD-ELMAN Electromyographic Signals
      WANG Wenmiao
      Computer Science. 2024, 51 (3): 257-264.  doi:10.11896/jsjkx.231000040
      Abstract ( 17 )   PDF(2871KB) ( 26 )   
      References | Related Articles | Metrics
      Surface electromyography(sEMG) signals are generated in advance of human movements and are commonly used to predict human behavior and motor intentions.However,due to its inherent non-stationary and time-varying characteristics,it is difficult to accurately predict changes in the angle of human lower limb.This paper presents a VMD-ELMAN angle fitting algorithm for muscle selection of human lower limb muscles for three movements:normal walking,ascending stairs and descending stairs.This algorithm improves the accuracy of surface electromyography signal angle prediction,enhances the real-time perfor-mance of angle prediction,and provides an effective solution for improving human-machine integration with exoskeleton devices.The Experimental results show that compared to common angle fitting algorithms,the proposed algorithm is less time-consuming.Among the three common movements,the highest accuracy of the hip joint angle prediction value RMSE is 0.578 9,and the knee joint angle prediction value RMSE is within 0.2.Its prediction accuracy is superior to common models,and the model has strong robustness.
      Online Electric Vehicle Charging Algorithm Based on Carbon Peak Constraint
      CAO Yongsheng, LIU Yang, WANG Yongquan, XIA Tian
      Computer Science. 2024, 51 (3): 265-270.  doi:10.11896/jsjkx.230800051
      Abstract ( 13 )   PDF(1727KB) ( 31 )   
      References | Related Articles | Metrics
      With the increasing number of electric vehicles(EVs),EV charging significantly increases the total load of the community,greatly increases the carbon emissions of the community,brings great instability to the community power grid,and reduces the power quality of the community.This paper studies the problem of scheduling EV charging based on the constraints of carbon peak when the arrival time,departure time,and charging demand of EVs are not known in advance.First,we formulate and study the problem of charging EVs without knowing future information.Aiming to address the uncertainty of EV charging behavior,we propose an algorithm for intelligent charging carbon emissions using the actor-critic approach,which learns the optimal strategy for EV charging through continuous charging instead of using a discrete approximation of carbon emissions.Simulation results demonstrate that compared with the online charging algorithm and the AEM energy management algorithm,the proposed algorithm can reduce the expected cost by 24.03% and 21.49%.
      Decentralized Federated Continual Learning Method Combined with Meta-learning
      HUANG Nan, LI Dongdong, YAO Jia, WANG Zhe
      Computer Science. 2024, 51 (3): 271-279.  doi:10.11896/jsjkx.230100125
      Abstract ( 14 )   PDF(3796KB) ( 27 )   
      References | Related Articles | Metrics
      For the problems of continual learning and data security in federated continual scenarios,a decentralized federated continual learning framework combined with meta-learning is constructed.First,in order to solve the problem of catastrophic forgetting in incremental scenarios,an incremental meta-learning method based nearest mean-of-exemplars replaying called NMR-cMAML is proposed.Then,to solve the problem of privacy security in federated continual scenarios,a decentralized federated continual framework based on peer-to-peer network architecture is designed,which is different from the center architecture based on server-client.Each client in the decentralized framework adopts NMR-cMAML to learn the continuous tasks incrementally,and the effective knowledge migration between clients is realized by sharing the meta-learning model in the federal communication process.Finally,experiments are conducted on image data sets(Cifar100 and Imagenet50) to verify that the proposed method improves the data privacy security of the system and improves the local performance of the client.
      Computer Network
      Multi-objective Optimization of D2D Collaborative MEC Based on Improved NSGA-III
      WANG Zhihong, WANG Gaocai, ZHAO Qifei
      Computer Science. 2024, 51 (3): 280-288.  doi:10.11896/jsjkx.221100250
      Abstract ( 12 )   PDF(2479KB) ( 22 )   
      References | Related Articles | Metrics
      In the current mobile edge computing(MEC),since tasks are directly uploaded to the MEC server for execution,there are problems such as high computing pressure on the edge server and insufficient utilization of resources on idle mobile devices.Using idle devices in the edge network for collaborative computing can realize rational utilization of user's idle resources and enhance the computing capacity of MEC.Therefore,a device-to- device(D2D) collaborative MEC for partial offloading(DCM-PO) is proposed.In this model,in addition to local computing and MEC server computing,part of the tasks can be uploaded to idle D2D devices for auxiliary computing.First,a multi-objective optimization problem is established to minimize the delay,energy consumption and cost of the edge network.Then,the non-dominated sorting genetic algorithm III(NSGA-III) is improved in the aspects of multi-chromosome mixed coding,adaptive crossover rate and mutation rate,so that it is suitable for solving the multi-objective optimization problem in the DCM-PO.Finally,simulation results show that,compared with the baseline MEC,the DCM-PO has advantages in several performance indicators.
      Green Energy-saving Routing Framework Based on Link Correlation Model
      WANG Ling, JIN Zikun, WU Yong, GENG Haijun
      Computer Science. 2024, 51 (3): 289-299.  doi:10.11896/jsjkx.230800103
      Abstract ( 13 )   PDF(2657KB) ( 24 )   
      References | Related Articles | Metrics
      With the rapid development of information technology,the scale of the Internet is increasing.At the same time,the energy consumption of the network is rising.In order to reduce network energy consumption,the industry generally adopts the method of closing the link with low link utilization.However,the current network energy-saving scheme can not effectively ba-lance the trade-off among energy-saving rate,computational overhead and path stretch.In order to solve the above problems,this paper proposes a green energy-saving routing framework based on link correlation model.The framework supports different link correlation models.It only needs the network topology,not the real-time traffic matrix in the network,so it is easier to deploy in the actual network.Based on the green energy-saving routing framework based on link correlation model,this paper implements four different energy-saving green routing algorithms:LRC(link row correlation),LCC(link column correlation),LRCC(link row column correlation) and LBC(link betweenness correlation).Experimental results show that in the real topology published by The Internet Topology Zoo and the topology generated by Brite simulation,the average energy saving rate of LRC,LCC,LRCC and LBC is 12.65% and 7.17% higher than that of DLF algorithm,and their average path stretch under the real topology and simulated topology is 3.00% and 13.75% lower than that of DLF algorithm.
      Efficient Routing Algorithm Based on Virtual Currency Transaction in DTN
      CUI Jianqun, LIU Shan, CHANG Yanan, LIU Qiangqiang, WU Qingcheng
      Computer Science. 2024, 51 (3): 300-308.  doi:10.11896/jsjkx.221200135
      Abstract ( 11 )   PDF(2762KB) ( 20 )   
      References | Related Articles | Metrics
      Due to the characteristics of intermittent connection of the delay tolerant network,as well as the limited resources such as the node's own cache and energy,the nodes in the DTN tend to show a certain degree of selfishness.The existence of selfish nodes may increase the network overhead and reduce the successful delivery rate of messages.In order to promote selfish nodes to participate in cooperation,an efficient routing algorithm PVCT(efficient routing algorithm based on virtual currency transaction in DTN) is proposed,which combines the small world characteristics of the delay tolerant network to improve the efficiency of the routing algorithm.The algorithm uses the virtual currency transaction mode,and prices according to the basic attributes,location attributes,social attributes,etc.of the node.The node gives the corresponding quotation according to the designed price function,and uses the price function to reasonably allocate the number of message copies.In PVCT strategy,nodes are divided into normal nodes and selfish nodes according to their judgments.When the number of hops of messages is less than or equal to two hops,they are forwarded according to the probability routing strategy.On the contrary,when the number of hops of the message is more than two hops,if a selfish node is encountered,the routing algorithm of the virtual currency transaction is executed.If the bid of the message carrying node is higher than the price of the forwarding node,the transaction will be conducted and the respective revenue status will be updated.Otherwise,entering the secondary price adjustment stage to coordinate the virtual quotation of both parties.Simulation results show that PVCT routing algorithm can better promote message forwarding in DTN,thus improving the overall performance of the network.
      Computation Offloading with Wardrop Routing Game in Multi-UAV-aided MEC Environment
      WANG Xinlong, LIN Bing, CHEN Xing
      Computer Science. 2024, 51 (3): 309-316.  doi:10.11896/jsjkx.221100242
      Abstract ( 16 )   PDF(2717KB) ( 18 )   
      References | Related Articles | Metrics
      The combination of Unmanned aerial vehicles(UAVs) and multi-access edge computing(MEC) technology breaks the limitations of traditional terrestrial communications,which has become a significant approach to solve the tasks offloading pro-blem in MEC.Due to the limited computing resources and energy that a single UAV can provide,the tasks offloading problem in a multi-UAV-assisted MEC environment is considered to cope with the growing network scale.Based on the problem definition,to obtain the offloading strategies in the equilibrium and optimal states and analyze the gap between them quantitatively,the tasks offloading process can be viewed as a Wardrop routing game on parallel links with player-specific latency functions.Since the equilibrium solution is difficult to compute,a new potential function is introduced to convert the equilibrium problem into a minimization problem of potential function.Simultaneously,the Frank-Wolfe algorithm is used to obtain the equilibrium and the optimal offloading strategies finally.At each iteration of this algorithm,the objective function is linearized,and the feasible direction is thus obtained by solving the linear programming,along which a one-dimensional search is performed in the feasible domain.Simulation experiments verify that the equilibrium offloading strategy based on the Wardrop routing game on parallel links can effectively reduce the model's total cost compared with other benchmark methods,and the ratio between the total costs caused by the equilibrium and optimal offloading strategies is about 1.
      Improved Beluga Whale Optimization for RFID Network Planning
      CHEN Yijun, ZHENG Jiali, LI Zhiqian, ZHANG Jiangbo, ZHU Xinghong
      Computer Science. 2024, 51 (3): 317-325.  doi:10.11896/jsjkx.230300019
      Abstract ( 15 )   PDF(3536KB) ( 21 )   
      References | Related Articles | Metrics
      With the development of radio frequency identification(RFID) technology,the demand for its application is getting higher and higher,and the research in reader deployment is gradually deepening.In order to solve the RFID reader location planning problem in the defined area,a mathematical optimization model is established with the objectives of tag coverage,collision interference between readers and load balancing in the delimited area,and an improved beluga whale optimization is proposed on the basis of the beluga whale optimization.Firstly,to address the shortcomings of the standard beluga whale optimization,which is easy to fall into the local optimum and lose the suboptimal solution,an update elite group mechanism is proposed.Secondly,to enhance the exploration capability of the algorithm,an opposition-based learning strategy is added,Finally,the algorithm is applied to solve the RFID network planning problem.By placing different numbers of clusters and randomly distributed tags in a certain environment,the improved beluga whale optimization is compared with the particle swarm algorithm,the gray wolf algorithm and the standard beluga whale optimization and the results are derived.Simulation results show that the performance of the improved beluga whale optimization improves on average 21.1% over the particle swarm optimization,28.5% over the grey wolf optimizer,and 3.3% over the beluga whale optimization in the same environment,indicating that the algorithm has better performance than the other three algorithms in terms of search accuracy,then,the effectiveness and feasibility of the application are verified by reader optimization deployment tests.
      Information Security
      Cryptographic Protocol Reverse Method Based on Information Entropy and Closed Frequent Sequences
      LIANG Chen, HONG Zheng, WU Lifa, JI Qingbing
      Computer Science. 2024, 51 (3): 326-334.  doi:10.11896/jsjkx.221200147
      Abstract ( 20 )   PDF(2810KB) ( 26 )   
      References | Related Articles | Metrics
      Unknown cryptographic protocols are widely used for the secure transmission of sensitive information,and reversing cryptographic protocol is of great significance to both attackers and defenders.In order to efficiently reverse complex cryptographic protocols,a cryptographic protocol reverse method based on information entropy and closed frequent sequences is proposed.The information entropy is used to distinguish the plaintext and ciphertext,and the closed frequent sequences mined by BIDE algorithm are used to identify dynamic fields and static fields in the messages.A length field identification algorithm is proposed.It slices the message,and compares the sliced field values with the set of length field values to achieve various forms of length field recognition in cryptographic protocols.Heuristic strategies are proposed to recognize the semantics of key fields including the fields specific to cryptographic protocols such as encryption suites and encryption algorithms.Experimental results show that the method can effectively identity fields and extract the formats of cryptographic protocols,outperforms the existing me-thods in various length fields identification and semantic recognition of key fields specific to cryptographic protocols as well.
      Blockchain Coin Mixing Scheme Based on Homomorphic Encryption
      WANG Dong, LI Zheng, XIAO Bingbing
      Computer Science. 2024, 51 (3): 335-339.  doi:10.11896/jsjkx.230100059
      Abstract ( 12 )   PDF(1347KB) ( 19 )   
      References | Related Articles | Metrics
      Coin mixing is important for protecting transaction privacy and realizing transaction unlinkability.However,hundreds of bytes of space overhead is necessary because of its verification process with pedersen commitment,which severely reduces its usability.A new coin mixing scheme is proposed by using SM2 algorithm,homomorphic encryption and stealth address technology in this paper.The on-chain transaction information is completely hide by using EC-ElGamal partially homomorphic encryption technology to encrypt the transaction value.Then the confidential transaction is sent to one-time stealth addresses after twice ve-rification and once re-randomization,thus breaking the connection between the payer and payee of the transaction to achieve unlinkability and untraceability of the transaction.This scheme can severely increase the privacy degree of transaction and transaction per second(TPS) while 82.25% reduction in the size of one transaction is achieved.At the same time,it enhances the resis-tance to analysis attacks,key replay attacks and sybil attacks.
      Dynamic Searchable Symmetric Encryption Based on Protected Search Mode of Updatable Encryption
      XU Chengzhi, XU Lei, XU Chungen
      Computer Science. 2024, 51 (3): 340-350.  doi:10.11896/jsjkx.230100016
      Abstract ( 14 )   PDF(1709KB) ( 22 )   
      References | Related Articles | Metrics
      Dynamic searchable symmetric encryption(DSSE) technology,as an extension of static searchable encryption,has attracted much attention because it solves the problem of secure retrieval over encrypted data and supports data dynamicity.For practicality concerns,most current DSSE schemes leak extra information(e.g.,search patterns and access patterns) to fast search.Recent studies show that this leaked information poses serious security problems,the adversary with background know-ledge of the database may exploit the leaked information to recover the query or reconstruct the database.Since this information reveals along with the query process,scholars propose to refresh the encrypted database after the query to reduce the above potential risks.However,this approach leads to huge client-side communication,storage,and computation overheads.Because the client needs to download the results locally,decrypt them,re-encrypt them and finally upload them to the cloud.To address this problem,this paper proposes a new updatable DSSE scheme that hides all the above information including access pattern,search pattern.The scheme can update data directly at the server side without disclosing data privacy,thus reducing the communication overhead of traditional update methods of the client side.The security analysis shows that this scheme can hide the search pattern effectively.In addition,the communication cost of the proposed scheme is also significantly degraded when compared with the traditional scheme that executes ciphertext refreshing by the client.For example,in the case of keywords matching 100 documents,compared with downloading to local re-encryption and retransmission,the communication overhead of this scheme is reduced by 70.92%.
      CheatKD:Knowledge Distillation Backdoor Attack Method Based on Poisoned Neuronal Assimilation
      CHEN Jinyin, LI Xiao, JIN Haibo, CHEN Ruoxi, ZHENG Haibin, LI Hu
      Computer Science. 2024, 51 (3): 351-359.  doi:10.11896/jsjkx.221200035
      Abstract ( 10 )   PDF(2366KB) ( 24 )   
      References | Related Articles | Metrics
      With the continuous performance improvement of deep neural networks(DNNs),their parameter scale is also growing sharply,which hinders the deployment and application of DNNs on edge devices.To solve this problem,researchers propose knowledge distillation(KD).Small student models with high performance can be generated from KD,by learning the “dark knowledge” of large teacher models,realizing easy deployment of DNNs on edge devices.However,in the actual scenario,users often download large models from public model repositories,which lacks the guarantee of security.This may pose a severe threat to KD tasks.This paper proposes a backdoor attack for feature KD,named CheatKD,whose backdoor,embedded in the teacher model,can be retained and transferred to the student model during KD,and then indirectly poison the student model.Specifically,in the process of training the teacher model,CheatKD initializes a random trigger and optimizes it to control the activation values of some certain neurons of a particular distillation layer in the teacher model(i.e.,poisoned neuron),making their activation va-lues fixed to enable poisoned neuronal assimilation.As the result,the teacher model is backdoored while this backdoor can resist to KD filtration and be transferred to the student model.Extensive experiment on four datasets and six model pairs have verified that CheatKD achieves an average attack success rate of 85.7%.Besides,it has good generality for various distillation methods.
      Optimal Penetration Path Generation Based on Maximum Entropy Reinforcement Learning
      WANG Yan, WANG Tianjing, SHEN Hang, BAI Guangwei
      Computer Science. 2024, 51 (3): 360-367.  doi:10.11896/jsjkx.221200104
      Abstract ( 16 )   PDF(3454KB) ( 24 )   
      References | Related Articles | Metrics
      Analyzing intrusion intentions and penetration behaviors from the attackers' perspective is of great significance for guiding network security defense.However,most existing penetration paths are constructed based on the instantaneous network environment,resulting in reduced reference value.Aiming at this problem,this paper proposes an optimal penetration path generation method based on maximum entropy reinforcement learning,which can capture the approximate optimal behavior of multiple modes in the form of exploration under dynamic network environments.Firstly,the penetration process is modeled according to the attack graph and the vulnerability score,and the threat degree of the penetration behavior is described by quantifying the attack benefits.Then,considering the complexity of the intrusion behavior,a soft Q-learning method based on the maximum entropy model is developed.The stability of the penetration path is ensured by controlling the entropy value and the importance of the reward.Finally,the method is applied to a dynamic environment to generate a highly available penetration path.Simulation experimental results show that,compared with the existing baseline methods based on reinforcement learning,the proposed method has more robust environmental adaptability and can generate higher-yielding penetration paths at a lower cost.
      Census Associated Multiple Attributes Data Release Based on Differential Privacy
      YOU Feifu, CAI Jianping, SUN Lan
      Computer Science. 2024, 51 (3): 368-377.  doi:10.11896/jsjkx.230100013
      Abstract ( 17 )   PDF(5130KB) ( 29 )   
      References | Related Articles | Metrics
      The release of unprotected census statistics carries the risk of revealing residents' personal privacy information.Census data protection solutions based on differential privacy have received substantial attention from researchers.Existing methods address the consistency constraint among geographic regions of census statistics,but associated multi-attribute data with more complex hierarchical consistency constraints face the challenge of being unable to build in a single hierarchical tree under existing methods.In this paper,we propose a differentially privacy method for optimally consistent release of associated multiple attributes statistics within census regions,which can achieve efficient release of statistics with complex consistency constraints.Firstly,the consistency constraints among the complex associated multiple attributes are divided into relatively independent and easily solved multiple consistency constraints.Then,based on the structural characteristics of the census associated multiple attributes data,mathematical analysis is used to further optimize the efficiency based on the existing methods.Finally,the optimal consistent release is achieved by combining the approximation method of the multiple consistency constraints problem.Experiments on real census datasets and synthetic datasets show that the proposed method can outperform similar methods in efficiency performance by one to two orders of magnitude while maintaining the same accuracy as similar methods.
  • Computer Science
    NCTCS2017 (0)
    Network & Communication (193)
    Information Security (625)
    Software & Database Technology (3)
    Artificial Intelligence (727)
    Surveys (52)
    Graphics, Image & Pattern Recognition (58)
    WISA2017 (1)
    WISA2018 (1)
    NASAC 2017 (13)
    CGCKD 2018 (12)
    Netword & Communication (4)
    Graphics, Image & Pattern Recognition (7)
    CCF Big Data 2017 (7)
    NCIS 2017 (5)
    Graphics, Image & Pattem Recognition (10)
    Interdiscipline & Frontier (95)
    Review (33)
    Intelligent Computing (158)
    Pattern Recognition & Image Processing (82)
    Big Date & Date Mining (21)
    Interdiscipline & Application (207)
    ChinaMM 2017 (12)
    CCDM2018 (14)
    Graphics ,Image & Pattern Recognition (45)
    Pattem Recognition & Image Processing (25)
    Big Data & Data Mining (42)
    Surverys (5)
    NDBC 2018 (8)
    Graphics,Image & Pattern Recognition (11)
    Database & Big Data & Data Science (362)
    Computer Graphics & Multimedia (420)
    Intelligent Software Engineering (17)
    Databωe & Big Data & Data Science (14)
    Computer Software (68)
    Mobile Crowd Sensing and Computing (9)
    Software Engineering (11)
    Intelligent Edge Computing (14)
    New Distributed Computing Technologies and Systems (9)
    Image Processing & Multimedia Technology (166)
    Artificial Intelligence Security (12)
    Human-Machine Interaction (5)
    Intelligent Data Governance Technologies and Systems (14)
    Blockchain Technology (18)
    Invited Article (2)
    Multilingual Computing Advanced Technology (11)
    Novel Distributed Computing Technology and System (11)
    Survys (0)
    Network & Communication (0)
    Graphics, Image & Patten Recognition (0)
    Interdiscipline & Frontier (4)
    HPC China 2018 (6)
    Data Science (18)
    Advances on Multimedia Technology (11)
    Special Issue of Social Computing Based Interdisciplinary Integration (13)
    Smart IoT Technologies and Applications Empowered by 6G (10)
    Smart Healthcare (10)
    Software & Database Technology (48)
    ChinaMM2018 (14)
    ChinaMM2019 (0)
    Intelligent Mobile Authentication (10)
    Big Data & Data Science (167)
    Network & Cornmunication (4)
    Theoretical Computer Science (10)
    Computer Vision: Theory and Application (16)
    Software Engineering & Database Technology (8)
    Computer Architecture (26)
    Discipline Construction (2)
    High Performance Computing (22)
    Computer Science Theory (19)
    Computer Network (243)
    Computer Graphics& Multimedia (10)
    High Perfonnance Computing (10)
    Federated Leaming (11)
    Edge Intelligent Collaboration Technology and Frontier Applications (10)
    Database & Big Data & Data Science (36)
    Computer Graphics & Multimedia (32)
    Explainable AI (11)
    Database & Big Data & Data Science (9)
    Software & Interdiscipline (16)
    Data Security (11)
    Granular Computing & Knowledge Discovery (10)
    Big Data & Data Science (37)
    Network & Communication (13)
    Special Issue on the 50th Anniversary of Computer Science (1)
    Special Issue on the 54th Anniversary of Computer Science (1)
    Special Issue on the 55th Anniversary of Computer Science (1)
    Special Issue on the 56th Anniversary of Computer Science (1)
    Special Issue on the 58th Anniversary of Computer Science (1)
    Discipline Frontier (2)
    Computer Networ (14)
    Information Security Protection in New Computing Mode (6)
    Special Issue of Knowledge Engineering Enabled By Knowledge Graph: Theory, Technology and System (11)
    Computer Graphics & Multimedia (0)
    Granular Computing & Knowledge Discovery (6)
    Interdiscipline & Frontier (0)
    Computer Software & Architecture (8)
    Interdiscipline & Application (14)
    Image Processing & Multimedia Technology (36)
    Special Issue on the 51th Anniversary of Computer Science (1)
    Special Issue on the 52th Anniversary of Computer Science (1)
    Special Issue on the 53th Anniversary of Computer Science (1)
    Special Issue on the 57th Anniversary of Computer Science (1)
    Special Issue on the 59th Anniversary of Computer Science (1)
    Contents (9)

  • 2024 Vol. 51 No. 2

  • 2024 Vol. 51 No. 1

  • 2023 Vol. 50 No. 12

  • 2023 Vol. 50 No. 11A
  • 2024,51 (2) 
  • 2024,51 (1) 
  • 2023,50 (12) 
  • 2023,50 (11A) 
  • 2023,50 (11) 
  • 2023,50 (10) 
  • 2023,50 (9) 
  • 2023,50 (8) 
  • 2023,50 (7) 
  • 2023,50 (6A) 
  • 2023,50 (6) 
  • 2023,50 (5) 
  • 2023,50 (4) 
  • 2023,50 (3) 
  • 2023,50 (2) 
  • 2023,50 (1) 

  • More>>
  • Web Application Page Element Recognition and Visual Script Generation Based on Machine Vision (429)
    LI Zi-dong, YAO Yi-fei, WANG Wei-wei, ZHAO Rui-lian
    Computer Science. 2022, No.11:65-75
    Abstract (429) PDF (2624KB) (11684)
    Optimization Method of Streaming Storage Based on GCC Compiler (1003)
    GAO Xiu-wu, HUANG Liang-ming, JIANG Jun
    Computer Science. 2022, No.11:76-82
    Abstract (1003) PDF (2713KB) (11665)
    Polynomial Time Algorithm for Hamilton Circuit Problem (5788)
    JIANG Xin-wen
    Computer Science. 2020, No.7:8-20
    Abstract (5788) PDF (1760KB) (11602)
    Review of Time Series Prediction Methods (3566)
    YANG Hai-min, PAN Zhi-song, BAI Wei
    Computer Science. 2019, No.1:21-28
    Abstract (3566) PDF (1294KB) (11417)
    Patch Validation Approach Combining Doc2Vec and BERT Embedding Technologies (312)
    HUANG Ying, JIANG Shu-juan, JIANG Ting-ting
    Computer Science. 2022, No.11:83-89
    Abstract (312) PDF (2492KB) (11070)
    Semantic Restoration and Automatic Transplant for ROP Exploit Script (275)
    SHI Rui-heng, ZHU Yun-cong, ZHAO Yi-ru, ZHAO Lei
    Computer Science. 2022, No.11:49-54
    Abstract (275) PDF (2661KB) (10782)
    Study on Effectiveness of Quality Objectives and Non-quality Objectives for Automated Software Refactoring (295)
    GUO Ya-lin, LI Xiao-chen, REN Zhi-lei, JIANG He
    Computer Science. 2022, No.11:55-64
    Abstract (295) PDF (3409KB) (10738)
    Decision Tree Algorithm-based API Misuse Detection (647)
    LI Kang-le, REN Zhi-lei, ZHOU Zhi-de, JIANG He
    Computer Science. 2022, No.11:30-38
    Abstract (647) PDF (3144KB) (10672)
    AutoUnit:Automatic Test Generation Based on Active Learning and Prediction Guidance (649)
    ZHANG Da-lin, ZHANG Zhe-wei, WANG Nan, LIU Ji-qiang
    Computer Science. 2022, No.11:39-48
    Abstract (649) PDF (2609KB) (10585)
    Research and Progress on Bug Report-oriented Bug Localization Techniques (614)
    NI Zhen, LI Bin, SUN Xiao-bing, LI Bi-xin, ZHU Cheng
    Computer Science. 2022, No.11:8-23
    Abstract (614) PDF (2280KB) (10308)
    Study on Integration Test Order Generation Algorithm for SOA (516)
    ZHANG Bing-qing, FEI Qi, WANG Yi-chen, Yang Zhao
    Computer Science. 2022, No.11:24-29
    Abstract (516) PDF (1866KB) (10055)
    Studies on Community Question Answering-A Survey (272)
    ZHANG Zhong-feng,LI Qiu-dan
    Computer Science. 2010, No.11:19-23
    Abstract (272) PDF (551KB) (9915)
    Research Progress and Challenge of Programming by Examples (529)
    YAN Qian-yu, LI Yi, PENG Xin
    Computer Science. 2022, No.11:1-7
    Abstract (529) PDF (1921KB) (9311)
    Survey of Distributed Machine Learning Platforms and Algorithms (1710)
    SHU Na,LIU Bo,LIN Wei-wei,LI Peng-fei
    Computer Science. 2019, No.3:9-18
    Abstract (1710) PDF (1744KB) (7843)
    Survey of Cloud-edge Collaboration (1550)
    CHEN Yu-ping, LIU Bo, LIN Wei-wei, CHENG Hui-wen
    Computer Science. 2021, No.3:259-268
    Abstract (1550) PDF (1593KB) (6826)
    Survey of Fuzz Testing Technology (1055)
    ZHANG Xiong and LI Zhou-jun
    Computer Science. 2016, No.5:1-8
    Abstract (1055) PDF (833KB) (6693)
    Overview on Multi-agent Reinforcement Learning (2816)
    DU Wei, DING Shi-fei
    Computer Science. 2019, No.8:1-8
    Abstract (2816) PDF (1381KB) (6656)
    Methods in Adversarial Intelligent Game:A Holistic Comparative Analysis from Perspective of Game Theory and Reinforcement Learning (2417)
    YUAN Wei-lin, LUO Jun-ren, LU Li-na, CHEN Jia-xing, ZHANG Wan-peng, CHEN Jing
    Computer Science. 2022, No.8:191-204
    Abstract (2417) PDF (4699KB) (6335)
    Survey on Attacks and Defenses in Federated Learning (888)
    CHEN Ming-xin, ZHANG Jun-bo, LI Tian-rui
    Computer Science. 2022, No.7:310-323
    Abstract (888) PDF (4808KB) (5941)
    Survey of Natural Language Processing Pre-training Techniques (1010)
    LI Zhou-jun,FAN Yu,WU Xian-jie
    Computer Science. 2020, No.3:162-173
    Abstract (1010) PDF (1678KB) (5800)
Announcement
Subject