基于混合行为特征的流量识别技术研究与应用

发布时间：2018-05-03 02:38

本文选题：流量识别 + 行为特征　；参考：《北京邮电大学》2017年硕士论文

【摘要】：随着互联网技术的高速发展和信息时代的到来,互联网相关应用已经深深影响到到人类社会生活的方方面面。网络的大量使用造成了互联网数据的爆发式增长,加重了网络安全和网络监控任务的负担。而要保障网络安全,进行网络分析的前提即为对流量的识别,因此,在网络的入口处对网络流量进行识别是非常重要的。网络流量的准确识别不仅可以洞察到整个网络的运行状况,而且可以针对特定流量进行精确监管,做到既保证网络的高速运行,同时对网络攻击,网络风暴等行为做出预防。但是,随着业务需求的不断增加,各种新型网络应用得到了快速发展,这给现有的流量分类方法带来了严峻的挑战。传统流量识别技术网络发展的不同时期都发挥了巨大的作用,但在现在复杂的网络状况下都存在着不同的技术瓶颈。本文吸取了传统流量识别技术的经验教训,从图的角度出发,将流行为检测和复杂网络特性相结合,提出新的流量识别模型。本文的工作主要分为如下部分。首先,从流量的行为统计特征出发,研究机器学习的分类方法在流量识别领域的应用,抽象出流量在传播过程中的行为特征,并选取合适的行为特征。其次,研究了复杂网络,对其主要思想和算法进行深入分析,提出图的特征在流量识别领域的应用。然后,研究如何将流行为和图连接行为相结合,构建更全面的混合行为流量识别模型。最后,利用Spark大数据计算平台,实现高效准确的流量识别系统。新的流量识别模型相较于以往仅关注单一层面的流量特征的分类模型是一个有力的补充,能取得更好的分类效果。同时,利用Spark平台在大数据处理中的优势,使流量识别系统更适合对海量数据流量的处理。
[Abstract]:With the rapid development of Internet technology and the arrival of the information age, Internet-related applications have deeply affected all aspects of human social life. The extensive use of the network results in the explosive growth of Internet data and increases the burden of network security and network monitoring tasks. In order to ensure the network security, the premise of network analysis is to identify the traffic, so it is very important to identify the network traffic at the entrance of the network. The accurate identification of network traffic can not only insight into the operation of the whole network, but also can accurately monitor the specific traffic, not only to ensure the high-speed operation of the network, but also to prevent the network attacks, network storms and other behaviors. However, with the continuous increase of service requirements, various new network applications have been rapidly developed, which brings serious challenges to the existing traffic classification methods. The traditional traffic identification technology has played a great role in different periods of the development of the network, but in the current complex network conditions, there are different technical bottlenecks. This paper draws on the experience and lessons of traditional traffic identification technology, from the view of graph, combines the popular detection with the characteristics of complex network, and puts forward a new traffic identification model. The work of this paper is divided into the following parts. Firstly, based on the statistical characteristics of traffic behavior, this paper studies the application of machine learning classification method in traffic recognition, abstracts the behavior characteristics of traffic in the process of traffic propagation, and selects appropriate behavior characteristics. Secondly, the complex network is studied, its main ideas and algorithms are analyzed in depth, and the application of graph features in the field of traffic recognition is put forward. Then, we study how to combine popular behavior with graph join behavior to construct a more comprehensive mixed behavior traffic identification model. Finally, the Spark big data computing platform is used to realize an efficient and accurate flow identification system. The new traffic recognition model is a powerful supplement compared with the previous classification model which only focuses on the traffic characteristics of a single layer and can achieve better classification results. At the same time, using the advantage of Spark platform in big data processing, the traffic identification system is more suitable for mass data traffic processing.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.06

【相似文献】