Loading...
  • 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

Current Issue

    • 论文
      A novel multicast routing algorithm based on bubble flow control   
      LOU Hui,XIAO Canwen,DONG Dezun,PANG Zhengbin,LI Cunlu
      2015, 37(02): 191-198. doi:
      Abstract ( 128 )   PDF (1213KB) ( 296 )     

      Multicast communication contributes most of communicating amount in Chip Multiprocessors (CMPs),and plays a key role in the performance of parallel applications.The recently proposed multicast routing algorithms most lack the balance in resource utilization,thus resulting in the low utilization of the buffers in NetworkonChip (NoC).A new multicast routing algorithm based on bubble flow control is proposed,which makes full use of the buffers in NoCs and avoids deadlock by injecting bubbles into the network.Our evaluation shows that the proposal can improve the performance of multicast communication efficiently.Under synthetic traffic,the proposal can reduce the average network delay by 18.1% and improve the network saturation throughput by 16.7%,compared with the balanced adaptive multicast routing algorithm.

      CbRouter:A bidirectional NoC router  architecture using crossbar bypass  
      FANG Lei,DONG Dezun,WU Ji,XIA Jun,WANG Kefei
      2015, 37(02): 199-206. doi:
      Abstract ( 123 )   PDF (4592KB) ( 330 )     

      Network-on-Chip (NoC) provides efficient communication support for multicore architectures.Typically,a general NoC is built with unidirectional channels and has a low utilization ratio of links.In order to dynamically allocate the link bandwidth and efficiently use the link bandwidth resources,a new bidirectional channel scheduling algorithm is proposed.A router architecture supporting this bidirectional channel scheduling algorithm is also designed.This novel router architecture provides a crossbar bypass data channel to rapidly transmit data with minor modification of the original data path.Experimental results show that the network saturation throughput and the link utilization get a maximum improvement of 83.3% and 24.53% respectively in comparison to the conventional Mesh with unidirectional links.

      Effect of compression on Hadoop:A case study of
      improving I/O performance on Hadoop   
      XIANG Lihui,MIAO Li,ZHANG Dafang
      2015, 37(02): 207-212. doi:
      Abstract ( 132 )   PDF (1043KB) ( 244 )     

      Nowadays, the development of disk I/O never catches up with CPU according to the Moore’s law, and Network I/O is scarce, so I/O often becomes a bottleneck of data processing. Hadoop can store PBlevel data where I/O problem becomes more obvious. Compression is an important method to optimize I/O, which can reduce I/O load and speed up data transmission on disk and network. In Hadoop, the benefits of using compression have not been completely exploited. In this paper we first analyze the compression algorithms supported by Hadoop, then propose a strategy to help Hadoop users identify how to use compression and how to verify through experiment. By using compression, the performance of some Hadoop applications can be improved up to 65%.

      Design and implementation of an antiIP
      fragmentation-evasion technique  
      LIU Baochao,ZHANG Yi,ZHANG Bofeng
      2015, 37(02): 213-218. doi:
      Abstract ( 110 )   PDF (3757KB) ( 223 )     

      Analyzing the detection technology of current NIDS,and the mechanism of formation and reassembly of IP fragmentations,we find that the conventional NIDS detection methods can’t detect the attack signatures contained in IP fragmentation very well,which is due to different fragmentation treatment strategies of different systems.Besides,the results of end hosts can’t be deduced from the results of NIDS,so the inconsistent behaviors between NIDS and end hosts,which means there may exist attack signatures in IP fragmentation,can easily evade NIDS detection.Therefore,we propose an antiIP fragmentationevasion method by adding a TPE  in the front of NIDS by serial method,which presets rules for IP fragmentation.Experimental results show that our method can effectively resist the IP fragmentation attack by about 90%.

      Intelligent resource center framework
      for resource integration on Internet of Things  
      YANG Dongju1,LIN Shaofu2,ZHANG Cheng3
      2015, 37(02): 219-225. doi:
      Abstract ( 129 )   PDF (1899KB) ( 248 )     

      Through active sense,intelligent analysis and feedback control,Internet of Things enforces the identification,monitoring and management of objects,which is successfully applied in many areas.However,there exists some issues in the development, such as decentralization,disorder,lack of system programming,duplicate construction,etc.We review the history of Internet of Things, analyze its technology architecture and related work. Then,the framework of IRC4IOT (Intelligent Resource Center for Internet of Things) is proposed to provide fundamental support for enabling the resources sharing,improving the resources utilization,and solving redundant construction through the convergence,integration,analysis and rebuilding of resources.The key technologies of how to join sensor devices into resource center and how to integrate and manage the resources are discussed.Finally,the cases in agriculture Internet of Things demonstrate the effect of the proposed framework.  

      A realtime scheduling algorithm
      based on protection threshold for CPS   
      ZHOU Benhai,YAO Dapeng
      2015, 37(02): 226-230. doi:
      Abstract ( 117 )   PDF (1689KB) ( 258 )     

      CPS is a novel complex realtime distributed system consisting of computing,communication and control,in which computation process and physical process interacts and integrates in an open environment.In order to respond to the physical world information in real time, the system usually adopts preemptive scheduling methods to guarantee the implementation of crucial tasks before deadline.However,in the distributed environment,preemptive scheduling methods easily lead to frequent task switching, thus affecting the system’s real-time performance. We propose a realtime scheduling algorithm based on protection threshold to maximize the execution time of low priority tasks by establishing a protection threshold model which can reduce the frequency of task switching. Experimental results show that the proposed algorithm effectively reduces the frequency of task switching, thus improving the realtime performance of CPS system. 

      Survey on topic-focused crawlers 
      YU Juan,LIU Qiang
      2015, 37(02): 231-237. doi:
      Abstract ( 402 )   PDF (487KB) ( 53349 )     

      With the exponential growth of network information resources and the growing personalized demands of customers, topicfocused crawler emerges as the times require. Topicfocused crawlers are programs designed to download web pages which are relevant to specific topics. Using information gathered at running time, topicfocused crawlers explore the webs which follow promissory hyperlinks, and fetch only pages which appear to be relevant. The searching engine and corpus building based on topicfocused crawling have been widely used. We first define the goals and operating principles of focused crawling, comprehensively analyze the recent advances at home and abroad, and then compare the crawling strategies of various topicfocused crawlers as well as the advantages and disadvantages of related algorithms. Finally, we point out the future direction of topicfocused crawling.

      An access control scheme in cloud storage based on
      content extraction signature and attribute encryption  
      WANG Caifen,XU Ting,ZHANG Yulei,YANG Xiaodong
      2015, 37(02): 238-244. doi:
      Abstract ( 129 )   PDF (529KB) ( 301 )     

      Existing access control schemes in cloud storage can only send different messages according to users’ attributes, which leads to low utilization rate. In order to overcome this shortcoming, firstly, we design a content extraction signature scheme based on ElGamal; secondly, we propose a new access control scheme in cloud storage based on content extraction signature and attribute encryption system. The proposed scheme combines access control with attribute encryption and realizes finegrained access control at the same time. The scheme can send submessages to corresponding users depending on users’ attributes; it improves the utilization of the message and has better applicability.

      PbRED:An improved RED algorithm based on priority  
      JIANG Ming,LIU Feng
      2015, 37(02): 245-251. doi:
      Abstract ( 118 )   PDF (849KB) ( 232 )     

      As an important active

      queue management algorithm, Random Early Detection (RED) obtains a desirable throughput

      performance and keeps a stable queue length. However, when there are distinct priorities among

      different traffics, it cannot differentiate the QoS of the traffics. A novel improved RED

      algorithm, called PbRED, is proposed by adjusting the dropprobability based on priorities. By

      decreasing the dropprobability of higher priority and vice versa, PbRED provides

      differentiated QoS for traffics with different priorities. The simulation results show that

      besides maintaining a higher total throughput, PbRED can provide reasonable differentiated QoS

      for traffics with different priorities and can guarantee that the traffic with a higher

      priority gets a better throughput performance.Key words:active queue management;RED

      algorithm;priority;queue length;throughput

      Design and implementation of
      Lucene-based full-text retrieval system  
      ZHOU Jingcai1,HU Huaping1,2,YUE Hong1
      2015, 37(02): 252-256. doi:
      Abstract ( 161 )   PDF (2817KB) ( 298 )     

      With the continuous improvement of informationization, a highperformance, full-featured text search system, which can fast locate the matching records among massive data, has become a new research hotspot. Based on the analysis of the fundamentals of the fulltext retrieval techniques and the structure of Lucene system, we present a MVCpattern fulltext retrieval model and develop a retrieval system based on SSH framework and Lucene search engine. It has three contributions. Firstly this system optimizes the supported file formats, and adds PDF, HTML, and RTF along with TXT, Ms office documents into the search library. Secondly, it improves the Chinese words segmentation machine in efficiency and accuracy. Thirdly, it enhances humanmachine interaction and achieves a similar display function as Baidu and Google, which can highlight the search keywords. The practical application of this system demonstrates that it is efficient in creating indexes and can speed up search with much more relevant results.

      A hidden communication system
      in BT network based on plugin technology    
      BAO Xudong,YOU Chun,LI Yongjun
      2015, 37(02): 257-262. doi:
      Abstract ( 116 )   PDF (1696KB) ( 212 )     

      BT network has been widely used in the Internet. The nodes in BT network feature anonymity and freedom, which is more conducive to hide the identity and to create steganographic system. We used Lehmer code algorithm to construct a hidden communication system in BT network based on plugin technology. The function of this system is that a special subgroup of peers in a BT network can identify each other and communicate without suspicion among regular peers. In this process, the hidden information is transmitted and the BT network also works normally. The system makes steganographic channels secretly without exposing anyone's identity.

      A review of the influence maximization problem in social networks  
      YAN Qing,ZHANG Peng
      2015, 37(02): 263-268. doi:
      Abstract ( 160 )   PDF (721KB) ( 308 )     

      Influence maximization is a problem of finding the most influential set of nodes in a social network which can maximize the spread of influence. The influence maximization problem has been one of the hot topics in social networks over the past decade. The study is not only of theoretical significance, but also of promising applications. The basic principles and the background of the problem are introduced. We also discuss the current development status, existing problems, some diffusion models and several algorithms for solving these problems. Finally, some future research directions of the problem are proposed.

      Reducing trust in the PKG and anonymous
      identity-based encryption scheme in the standard model  
      MING Yang,HAN Juan,ZHANG Lin,ZHOU Jun
      2015, 37(02): 269-275. doi:
      Abstract ( 126 )   PDF (492KB) ( 245 )     

      Due to anonymity and key escrow problem in the IBE scheme, a reducing trust in the PKG and anonymous identitybased encryption scheme is proposed based on the scheme of Waters and the ideas of Goyal on composite order groups, which is secure without random oracles in the standard model. The analysis shows that the scheme satisfies anonymity and reduces trust in the PKG under the DBDH assumption and DL assumptions. It can not only protect receivers’ privacy, but effectively solve the key escrow problem.

      A node reusable virtual network embedding algorithm
      based on discrete particle swarm optimization  
      LIU Xiangdong,LIU Kui,WANG Cong
      2015, 37(02): 276-280. doi:
      Abstract ( 177 )   PDF (737KB) ( 223 )     

      One major challenge in network virtualization is the efficient mapping of virtual networks onto substrate networks, which is known as the Virtual Network Embedding (VNE) problem. It’s also a basic multitenancy technology in cloud computing environment. A node reusable VNE algorithm is proposed in this paper. Firstly, a resource allocation optimal model is presented in order to achieve high utilization ratio of substrate networks. Then leveraging the advantage of reusable technology which uses ram switch instead of network exchange, we design an enhanced initial position assignment mechanism to improve the computational efficiency. The whole discrete particle swarm optimization based algorithm is presented lastly. Simulation results show that, compared with existing PSO based approaches, the proposed algorithm can   support the same number of virtual networks on the same substrate networks at lower cost, and can further improve computational efficiency.Key words: 

      Research on properties of χ and
      θ mappings in SHA-3 permutation  
      WANG Gan,ZHANG Wenying
      2015, 37(02): 281-287. doi:
      Abstract ( 124 )   PDF (1543KB) ( 209 )     

      Keccak has attracted many scientists since  it was announced new standard of Hash function SHA3 last year. θ and χ are the two most important and complex mappings in Keccak permutation. In this paper we first give the boolean function expression of χ, and then make a differential distribution table of χ by computing output differences from 32 input differences. We also make an analysis of the distribution pattern of χ from the distribution table. Double Kernel guarantees that the differences will not be diffused when passing θ mapping. With regard to this, we propose a new Double Kernel search algorithm with lower complexity compared with the algorithm in [1]. In the end, theoretic derivation and experiments prove that the differences with Hamming Weight equal to or less than 4 do not exist.

      Design and implementation of a mobile
      learning system prototype on mobile Internet  
      DING Yizhe,WEI Renjia,LIU Bo,WU Zhenqiang
      2015, 37(02): 288-293. doi:
      Abstract ( 135 )   PDF (893KB) ( 233 )     

      We design and implement a mobile learning system prototype for quality courses and open courses.We also integrate  network teaching resources in the scenario of tripartite participation system.Combined with the features of current mobile intelligent terminals in the scenario of tripartite participation system,we propose a viable method which can reduce computing and communication resources waste.The correctness of the system is verified by using deterministic finite automata method.Experimental results show that the scheme can effectively reduce the work pressure of proxy service module and resource service module in system scenario,and improve the degree of user experience.

      A process metamodel to support portable reconfiguration 
      LI Xin,XU Yuankun
      2015, 37(02): 294-308. doi:
      Abstract ( 126 )   PDF (917KB) ( 196 )     

      The idea of separating process logic from business logic is promoted by workflow. However,it’s hard to be carried out in reality because of the ubiquitous dynamic relevance between process and business in the business process. Process control is usually coupled with business process in the form of program code,which makes process modeling complicated and brings a heavy burden to process reconfiguration.We analyze the Workflow Management Coalition (WFMC) process meta-model,which indicates that the limitations of the model in the expression of process logic cause the difficulty of supporting convenient reconfiguration of business process.So a new process meta-modelcalled  EventStateRule (ESR ) meta-model, is presented.Some elements are added to the proposed model,such as events,states and the rules with which the dynamic relevance between the process and the business can be normally expressed by means of process definition.The boundary between process logic and business logic becomes explicit under the framework of our model,which helps to separate process logic from business logic far better.Besides,rigid process and flexible process can be both modeled by using unified process definition.When process logic varies,process definition can be modified on different model layers according to variation requirements, thereby,portable reconfiguration of the process is implemented.

      A business rule generator based on mapping mechanism  
      LI Weichao,ZHOU Xingshe,YANG gang,YANG Yalei
      2015, 37(02): 309-313. doi:
      Abstract ( 115 )   PDF (671KB) ( 184 )     

      Because of the diversity of business rules languages,the information system based on rules engines is hard to upgrade to next releases or to migrate to different engines.In this paper we adopt a layered architecture and propose an engine-independent rule generator based-on mapping mechanism.We define an English-like syntax rules description model,design XMLbased intermediate rules,establish the mapping relations of transformation between intermediate rules and targeted rules,and implement mutual transformations among engine-special rules.

      New evaluation method for multiple expression programming design  
      ZHENG Qiusheng1,HE Pei1,2,3,LI Ji1
      2015, 37(02): 314-319. doi:
      Abstract ( 115 )   PDF (633KB) ( 177 )     

      Multiple Expression Programming(MEP)is an automatic programming approach widely used in many areas.According to the chromosome representation and population evolution rules,probably any gene of each chromosome may be used by some other genes of the current or other subsequent populations,which results in a high consumption of time and space resources due to duplicated counting.In light of this problem,a new evaluation approach is proposed.In the case that the traditional representation rules of MEP chromosome and the evolutionary way of the population are not changed,the new evaluation method can identify repeated genes accurately and effectively,thus making it possible to get rid of large duplicated computations, and improving evolutionary performance significantly.

      Instructional feedback strategy generation algorithm
      based on an emotion learning ontology  
      TAO Xiaomei,NIU Qinzhou
      2015, 37(02): 320-328. doi:
      Abstract ( 131 )   PDF (2656KB) ( 239 )     

      This paper studies how to solve the problem of emotion analysis and emotion feedback in instructional tutoring system. We construct an emotion learning ontology to describe the emotional concepts and the relationship among these concepts during the learning process, and we design an algorithm to select appropriate teaching feedback strategy on the basis of the emotion ontology.This algorithm can understand learners’situation and provide support without interaction with learners through questions and answers.The emotion ontology,a C programming ontology,and the algorithm are applied into an intelligent tutoring system which uses instructional videos as core materials.Application cases verify that the feedback strategy generation algorithm on the basis of the emotion learning ontology can provide appropriate cognitive feedback and emotional feedback to learners.