Dissertation > Excellent graduate degree dissertation topics show

P2P Traffic Identification Method

Author: ZhuChao
Tutor: YuMing
School: Dalian University of Technology
Course: Circuits and Systems
Keywords: P2P traffic identification Affinity propagation Semi-supervised clustering Machine Learning
CLC: TP393.06
Type: Master's thesis
Year: 2011
Downloads: 44
Quote: 1
Read: Download Dissertation


As the Internet continues to grow, P2P (Peer-to-Peer) technology with significantly better than traditional C / S mode network structures and efficient processing capacity of people's lives has brought great convenience. At the same time as people demand for Internet applications is rising, based on P2P file sharing technology, voice services and applications such as streaming media has been rapid development, but due to the characteristics of P2P makes its own structure for network management and maintenance brought a lot of difficulties, mainly reflected in the P2P applications occupy huge bandwidth resources, causing network congestion, thus affecting the normal use of other services. And P2P applications continues to try to evade regulation, the use of random ports, tunneling mechanism or application layer encryption and other means to make regular traffic identification methods can not effectively carry out its monitoring. Therefore, accurately and effectively identify P2P traffic as P2P traffic monitoring tasks currently facing the primary problem. Firstly, the current P2P traffic identification method already identified are analyzed, which specifically includes traffic based on port numbers to identify, based on deep packet inspection of traffic identification, traffic flow characteristics based recognition and machine learning-based traffic identification. As traffic identification based on machine learning methods is the current hot research field of traffic identification, this paper focuses on several commonly used machine learning algorithms were analyzed. Secondly, P2P traffic recognition feature selection problem, a related feature selection methods, and emphatically analyzed two typical feature selection algorithm in P2P traffic identification applicability, which is based correlation feature selection algorithm (Correlation -based Feature Selection, CFS) and consistency-based feature selection algorithm (Consistency-based Feature Selection, CON). The experimental results show that the use of CFS P2P traffic identification feature selection algorithm can ensure the accuracy of the premise recognition algorithm to shorten the training and recognition time. Finally, P2P traffic identification mark, when the proportion of training samples is low recognition rate of deterioration problems occur, we propose a semi-supervised strategy based on affinity propagation (Affinity Propagation, AP) clustering algorithm, the core idea is to use a small amount of Given the marked sample clustering as monitoring strategies and specific implementation steps are: (1) right - given the proportion of labeled training samples, and to promote its campaign to become first class through a representative point; (2) labeled samples through messaging cluster; (3) in accordance with the appropriate \The algorithm of two key parameters, the damping coefficient λ and bias parameter p, this paper also investigated through a series of experiments both recognition performance of the algorithm, and gives practical application of the recommended value. Experimental results show that the kernel estimation supervised Naive Bayes (Naive Bayes using Kernel density estimation, NBK) algorithm and semi-supervised K-means algorithm, this algorithm labeled training samples is less than 20% the proportion of cases has a higher recognition accuracy and lower false recognition rate. This means that when the algorithm for P2P traffic identification, the recognition performance can be guaranteed under the premise of reducing the training samples for marking work intensity, which makes the algorithm identified in the flow field of a high application value.

Related Dissertations

  1. Research on Peer-to-Peer Traffic Identification Algorithm Based on Cluster Analysis,TP393.02
  2. Based on Data Distribution Characteristics of Text Classification,TP391.1
  3. Research and Implement of Chinese Word Segment Techniques Based on the Conditional Random Field,TP391.1
  4. Semi-supervised Learning and Active Learning of Sentiment Classification Coupled with Domain Knowledge,TP181
  5. A Static Behavior-Based Method to Detect Malware on Android,TP309
  6. Supervision topic model research and application,TP391.1
  7. Distortion effects on image quality evaluation and classification,TP391.41
  8. Learning-based human motion synthesis inverse kinematics,TP391.41
  9. Content-based malicious code detection Research and Implementation,TP393.092
  10. Home Academic Information Extraction System,TP393.092
  11. Social network algorithm and its application in recommendation,TP391.3
  12. Application of Data Mining in Email Anti-Spam System,TP393.098
  13. Based on self-learning social relation extraction research,TP391.1
  14. Graph model based on statistical and machine learning algorithm and its application of certain,TP181
  15. Based on rough sets and SVM national defense Comprehensive Quality Assessment Methods,E075
  16. SVM Based on SIFT and scene classification,TP391.41
  17. News video story unit segmentation,TP391.41
  18. Machine learning based on sparse coding and image content recognition algorithm,TP391.41
  19. Random Forests Feature Selection,TP311.13
  20. Research on Techniques of Collecting Internet Traffic Ground Truth,TP393.06

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Computer network > General issues > Computer networks, test , run
© 2012 www.DissertationTopic.Net  Mobile