Algorithms for network anomaly and topology detection

IP网络越来越多地支持多样化的融合和关键任务网络服务，超出了最初设计的最大努力流量。这些服务的快速部署，加上IP架构和协议的分布式特性，导致IP网络复杂性明显上升。因此，对网络性能问题进行故障排除已成为一项耗时且代价高昂的工作。下一代性能监控系统将需要检查传统上不用于自动故障排除系统的新型网络数据。该项目将侧重于探索性研究和原型设计，以推进下一代监测系统两个关键领域的最新技术。首先，该项目将专注于开发新的基于netflow的异常检测算法。Netflow是一种网络协议，允许路由器将每流流量统计汇总信息导出给第三方网络和安全

IP networks increasingly support a diversity of converged and mission critical network services beyond the original best-effort traffic that they were designed to carry. Rapid deployment of such services coupled with the distributed nature of the IP architecture and protocols have led to a distinct rise in IP network complexity. As a result, troubleshooting network performance issues has become a time-consuming and costly endeavor. Next generation performance monitoring systems will need to examine new types of network data not traditionally utilized for automated troubleshooting systems. This project will focus on exploratory research and prototyping to advance the state of the art in two key areas for next generation monitoring systems. First, the project will focus on developing novel Netflow-based anomaly detection algorithms. Netflow is a network protocol that allows routers to export per-flow traffic statistics summary information to third party network and security