项目名称: 抽样环境下基于流记录的行为特征分析与多分类器识别模型研究
项目编号: No.U1504602
项目类型: 联合基金项目
立项/批准年度: 2016
项目学科: 无线电电子学、电信技术
项目作者: 董仕
作者单位: 周口师范学院
项目金额: 27万元
中文摘要: 网络流量识别对网络安全监控、网络管理等具有重要意义。本项目旨在探索抽样环境下基于流记录的流量特征分析及关键技术,拟建立基于流记录的流量识别模型及相应的识别算法。拟研究的内容及目标:研究网络流量行为数据捕获及分析模型,构建高精度基准数据集。综合考虑抽样环境下由于信息的缺失而带来的数据量的降低进而影响到属性选择的因素,分析不同抽样策略对特征属性的相关性的影响,给出抽样率与相关性之间的影响度关系模型,并设计一种适合抽样环境下的属性选择算法。综合考虑不同分类器之间的异构性及偏好性,和抽样策略所带来的影响,以提高抽样环境下的网络流量识别精度为目标,设计多分类器的流量识别模型,并给出多分类器融合的解决方案。本项目所提出的算法和模型不仅可以提高抽样环境下网络流量识别精度,还能适应于加密流量的识别问题。项目预期成果能有效推动网络流量识别技术的发展,对提高互联网安全监控和管理具有重要的理论及应用意义。
中文关键词: 网络管理;网络行为;属性选择;报文抽样;流量识别
英文摘要: Traffic identification is crucial to network security monitoring and network management.This project aims to explore the feature of flow record based traffic in packet sampling and key analysis techniques,and build the model of flow record based traffic identification and corresponding traffic identification algorithm.The intended content and objectives are:1)Investigate the capturing and analysis model of network traffic behavior and build a benchmark data set with high-precision;2)In consideration of information loss in packet sampling,analyze how sampling techniques impact the correlation of traffic features,provide the relationship between sampling rate and the influence degree of the correlation of traffic features,and design a feature selection algorithm tailed for packet sampling;3)To improve the precision of traffic identification in packet sampling,design a multiple classifier ensemble based traffic identification model which comprehensively considers the heterogeneity and preference of multiple classifiers and the impact of packet sampling.The proposed algorithms and model not only improve the precision of traffic identification in packet sampling,but also can identify encrypted traffic.The project is expected to promote the development of traffic identification technology,which has important theoretical and applied significance to the improvement of Internet security monitoring and management.
英文关键词: network management;network behavior;feature selection;packet sampling ;traffic identification