Title : Inferring Hidden Features of the Internet
Speaker: Gonca Gürsun (Boston University)
Date : Dec 26, 2012
Time : 13:40
Place : FENS L035
Please find below the abstract of the talk and a short bio of our speaker.
The Internet is a collection of thousands of autonomous systems (ASes), each controlled by profit-seeking businesses with different economic goals. The complicated relationships between these independent ASes result in a very large and complex AS-level Internet system whose global behavior is almost unknown. Each individual AS can only observe a small portion of the Internet traffic and routing state through measurements. Moreover, understanding the routing choices of ASes is a difficult task due to the size and complexity of millions of
In this talk, we address two statistical inference problems with the goal of extending the knowledge of AS-level Internet. First, we discuss the ability of individual ASes to extend their visibility of the AS-level Internet. We define two types of visibility: 1) the visibility of interdomain traffic volumes which corresponds to the knowledge about the amount of traffic exchanged between ASes, and 2) the visibility of interdomain routing state which corresponds to the knowledge about the set of routes that passes through an AS. Then, we show how ASes can infer the volume of traffic that does not pass through their networks and identify the set of routes that pass through their networks using various statistical inference techniques.
Second, we introduce a new approach to analysis of interdomain routing system designed to shed light on collective routing choices. We define a new path-based metric called Routing State Distance (RSD) and show how to use this metric as a tool for characterizing extensive dataset of routing paths. We show that such characterization uncovers some surprising patterns in the interdomain routing system.
Gonca Gürsun is last year PhD student in the Department of Computer Science at Boston University. She works in the field of computer networking under the supervision of her advisor, Prof. Mark Crovella. Gonca's research interests are in networking, statistical inference, and data mining. Her current research focuses on statistical inference techniques for network traffic analysis, with applications to network management, traffic engineering and business intelligence.