Research

• Inter Domain Routing

  My thesis deals primarily with the routing component of the Internet infrastructure. If I were asked to summarize my thesis in a paragraph, it would be the following

  The routing subsystem of the Internet architecture, born circa 1990, has not kept pace with the explosive growth of the Internet itself. The Internet, as it looks today (~317 million hosts, ~20k autonomous systems, messy--not to mention "meshy"--interconnection), looks very different from the structure that the routing system was designed to support. In my thesis, I study some significant shortcomings of BGP, the inter-domain routing protocol of the Internet, and suggest ways (both protocol-based and operational) in which it can be changed to better address the requirements of todays' Internet infrastructure.

  
  
  
  
  
  

• Internet Topology Discovery

  There are two distinct modes of interconnection for autonomous systems. In the first, an AS connects to a neighbor via a dedicated circuit. A second, more interesting mode is at Internet Exchange points, which are shared infrastructures. Identifying IXP's and the ASes that interconnect at them reveals interesting information about the evolution of the Internet AS graph and varying trends of interconnection and business arrangements in different parts of the world.
  We developed a methodology based on a combination of active probing, and archived BGP routing data to infer the existence of IXP's, which are often not directly observeable, and to identify the various ASes that interconnect at said IXPs.
  

• Network Infrastructure Services

  The Internet is today, a ubiquitous (information) infrastructure, playing a critical role in society. However, little has changed since its inception (architecture-wise), that reflects its pre-eminent role today. The implication is that the "best-effort" Internet architecure, which provides a basic connectivity service, is at odds with emerging services that are driving the growth of the Internet, which demand availability, robustness to failure, mobility and quality of service. Traditional means to bridge this divide is via the use of application specific overlay networks. overlay networks, while filling an immediate and specific need, introduce problems of their own.
  As a way to address these important requirements, we designed the framework for a unifying network substrate, named the Service Oriented Internet.
  Essentially, this platform serves as a flexible and unifying substrate to deliver services over the Internet. We use several applications to demonstrate the effectiveness of such an architecture in addressing the inadequacies of the Internet architecture.

• Network Security

  In the past, I have worked with the MINDS research group. The general research focus is in applying techniques from data mining and statistics to network security. I am also keenly interested in infrastructure based network security, i.e., in-network mechanisms to deal with DDoS attacks, IP spoofing, etc.