本文選題：在線流量分類　切入點(diǎn)：分層式流量區(qū)分系統(tǒng)　出處：《濟(jì)南大學(xué)》2014年碩士論文

【摘要】：隨著互聯(lián)網(wǎng)的普及，日益增長(zhǎng)的網(wǎng)絡(luò)流量和不斷出現(xiàn)的各種網(wǎng)絡(luò)新應(yīng)用大大增加了網(wǎng)絡(luò)的復(fù)雜性和管理難度，這嚴(yán)重威脅到網(wǎng)絡(luò)服務(wù)質(zhì)量和網(wǎng)絡(luò)安全，對(duì)網(wǎng)絡(luò)管理提出了巨大挑戰(zhàn)。在這種情況下，網(wǎng)絡(luò)流量分類具有良好的網(wǎng)絡(luò)應(yīng)用識(shí)別能力，并且能為網(wǎng)絡(luò)管理提供信息支持，從而逐漸成為學(xué)術(shù)界研究的熱點(diǎn)。網(wǎng)絡(luò)流量因具有實(shí)時(shí)、易逝和不可再現(xiàn)的特性，致使網(wǎng)絡(luò)流量分類必須擁有及時(shí)分類能力才更有實(shí)際意義，所以在線分類成為研究網(wǎng)絡(luò)流量分類的重點(diǎn)。 經(jīng)過(guò)研究各種網(wǎng)絡(luò)流量分類技術(shù)的原理和現(xiàn)狀，我們認(rèn)識(shí)到當(dāng)前網(wǎng)絡(luò)流量分類多以單點(diǎn)分類模式為主，其注重于分類準(zhǔn)確率改善，卻忽略了分類性能的不足而導(dǎo)致無(wú)法對(duì)日益增長(zhǎng)的園區(qū)網(wǎng)絡(luò)出口流量進(jìn)行在線分類的問(wèn)題。為了實(shí)現(xiàn)網(wǎng)絡(luò)流量在線分類，本文打破單點(diǎn)流量分類的模式，將分布式思想引入到流量分類中，提出了一種分層式網(wǎng)絡(luò)流量分類方法。這種方法是在園區(qū)網(wǎng)絡(luò)的低速鏈路部署多個(gè)分類節(jié)點(diǎn)進(jìn)行在線流量分類，并由數(shù)據(jù)中心對(duì)分類結(jié)果進(jìn)行匯總來(lái)實(shí)現(xiàn)整個(gè)網(wǎng)絡(luò)流量的在線分類。其主要特性：一是其避免了直接處理園區(qū)網(wǎng)絡(luò)出口流量，，克服了單點(diǎn)性能不足而無(wú)法實(shí)現(xiàn)在線分類的缺陷；二是分類器動(dòng)態(tài)加載機(jī)制能夠?qū)崿F(xiàn)對(duì)新應(yīng)用識(shí)別和分類器的訓(xùn)練及動(dòng)態(tài)更新，有效應(yīng)對(duì)了概念漂移引起的分類準(zhǔn)確率下降的現(xiàn)象。分層式網(wǎng)絡(luò)流量分類不失為一種解決在線流量分類的有效方法。 本文依據(jù)分層式網(wǎng)絡(luò)流量分類方法設(shè)計(jì)了一個(gè)分層式網(wǎng)絡(luò)流量分類模型，其包括中心節(jié)點(diǎn)、分類節(jié)點(diǎn)和Web集成管理等部分。中心節(jié)點(diǎn)和分類節(jié)點(diǎn)相輔相成的完成園區(qū)網(wǎng)絡(luò)的流量區(qū)分任務(wù)。Web集成管理是對(duì)分類系統(tǒng)管理的Web實(shí)現(xiàn)。在分類模型的實(shí)現(xiàn)過(guò)程設(shè)計(jì)提出了分層式交互協(xié)議和分類器分發(fā)管理方法；實(shí)現(xiàn)了基于半監(jiān)督學(xué)習(xí)的未知流量發(fā)現(xiàn)識(shí)別、分類器訓(xùn)練等功能；分別采用了LIBPCAP平臺(tái)和可視化技術(shù)用于進(jìn)行在線網(wǎng)絡(luò)流量的捕獲和結(jié)果顯示等。最后經(jīng)過(guò)實(shí)驗(yàn)和Web管理測(cè)試說(shuō)明了分層式網(wǎng)絡(luò)流量分類方法在線流量分類的可行性以及Web管理的實(shí)用性。
[Abstract]:With the popularity of the Internet, network traffic is increasing and the emergence of new network applications greatly increased network complexity and management difficulty, which is a serious threat to the network quality of service and network security, put forward a huge challenge to the network management. In this case, the network traffic classification has a good ability to identify network applications, and can provide information support for network management, which has gradually become the focus of academic research. The network traffic with real-time, perishable and irreproducible characteristics, resulting in network traffic classification and classification ability must have more practical significance, so online classification has become the focus of research on network traffic classification.
The principle and research status of all kinds of network traffic classification technology, we realize the network traffic classification by single point classification model, which focuses on improving the classification accuracy, but ignore the lack of classification performance and can lead to the online classification of campus network flow growing problem. In order to achieve network traffic online classification, this paper breaks the single point traffic classification model, will be distributed into traffic classification, a classification method of hierarchical network traffic. This is a method of online traffic classification in multi node classification low-speed link network deployment, and the data center on the classification results were aggregated to achieve the online classification of the whole network traffic. Its main characteristics: one is to avoid direct processing of campus network export flow, to overcome the shortcomings of single point performance can not be achieved The two is the defect of online classification; dynamic classifier training and dynamic loading mechanism can realize the new application in the identification and classification, to effectively deal with the concept of classification accuracy caused by the drift down phenomenon. Hierarchical network traffic classification is an effective method to solve the problem of online traffic classification.
Based on the hierarchical network traffic classification method to design a hierarchical network traffic classification model, which comprises a central node, node classification and Web integrated management. The central node and the node classification complement each other to complete the park network traffic differentiating task.Web integrated management is to realize the classification management system in the process of design and implementation of Web. The classification model proposed the method of hierarchical distribution management interaction protocol and classifier; to achieve a semi supervised learning based on identifying the unknown flow, classifier training and other functions; using LIBPCAP platform and visualization technology for online network traffic capture and display results. Finally, through experiment and Web test shows the feasibility of hierarchical management network traffic classification method for online traffic classification and Web management is practical.

【學(xué)位授予單位】：濟(jì)南大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類號(hào)】：TP393.06

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 張麗新;王家欽;趙雁南;楊澤紅;;機(jī)器學(xué)習(xí)中的特征選擇[J];計(jì)算機(jī)科學(xué);2004年11期

2 熊剛;孟姣;曹自剛;王勇;郭莉;方濱興;;網(wǎng)絡(luò)流量分類研究進(jìn)展與展望[J];集成技術(shù);2012年01期

本文編號(hào)：1680660