骨干通信网络的流行为特征分析与识别关键技术研究

项目名称： 骨干通信网络的流行为特征分析与识别关键技术研究

项目编号： No.61471101

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 无线电电子学、电信技术

项目作者： 胡光岷

作者单位： 电子科技大学

项目金额： 82万元

中文摘要： 在骨干通信网络的流行为特征分析基础上，开展与流行为相关的网络异常事件识别研究，是确保通信网络有效管理和安全的基础工作，是国内外学术界和工业界共同关注的前沿科学问题。项目围绕流行为特征分析与异常事件识别两项关键问题，对尚未解决的基础问题进行研究，获得全面准确的网络流行为及其变化特征，提高实际网络问题的解决能力。为应对流连接特征及其变化的复杂性，提出基于增强型复杂网络的流行为及演化特征分析方法；为应对流行为的时变特征，以时变信号分析为手段，提出流行为时变特征分析与特征参数提取方法；为应对异常事件识别的复杂性与告警信息的不完备性，以模糊分析为手段，提出基于演化特征的异常事件检测与模糊识别方法。主要创新工作有：网络连接图的成图方法、基于增强型复杂网络的流连接行为与演化特征分析方法、网络连接图的动态网络子图挖掘方法、流行为的时变特征分析与特征参数提取方法、告警信息不完备条件下的异常事件模糊识别方法。

中文关键词： 通信网络；流；行为特征；识别

英文摘要： Network anomaly detection researches based on the flow behavior feature analysis in backbone communication networks is the fundamental work which ensures the effective management and the security of communication network systems, and it is also a frontier scientific problem concerned by both academia and industry at home and abroad. Focusing on two key scientific issues, flow behavior analysis and anomaly detection, this project endeavors to research on unsolved basic problems, to get accurate and comprehensive features and patterns of both flow behavior and its changes in backbone communication networks, and to improve the ability of existing methods to solve practical problems in networking. In order to deal with the complexity of flow connection features and their changes, our project presents a method to analyze network flow behavior features based on the enhanced complex network; By adopting time-varying signal analysis methods, we present the time-varying feature analysis and feature extraction to counter the time-varying nature of network flow behavior; Based on the fuzzy analysis, we present the evolutionary-feature-based detection and fuzzy detection scheme to solve the complexity of anomaly detection and the incompleteness of alarm information. Our main innovative works include: network connection graph formation and simplification, flow connection behavior and evolutionary feature analysis based on enhanced complex network, dynamic sub-graph mining in network connection graphs, time-varying feature analysis and network flow behavior feature extraction, fuzzy anomaly detection under incomplete alarm information.

英文关键词： communication network;flow;behavior characteristic;recognition