[Busec] practice talk, tomorrow 10AM MCS137
goldbe at cs.bu.edu
Tue May 10 13:43:42 EDT 2011
I'm giving a new survey talk to a theory-ish audience in Israel next
week (bar ilan, hebrew university, and tel aviv), and I wanted to do a
practice talk before I go. The talk covers our new SIGCOMM paper
(with Phillipa) and some new work with Zhenming. 10AM in MCS137.
Please come by if you have time. Abstract below.
The economics of routing security.
Despite a decade of research, the problem of securing the global
Internet's routing system is far from solved. The plethora of viable
solutions for securing Internet routing (i.e. S*BGP) have made it
clear that this problem not one of technical feasibility, but rather
one of incentives. The Internet is a complex, decentralized system
consisting of thousands of autonomous systems with diverse economic
interests (e.g. backbone providers like AT&T, content providers like
Google, business networks like Bank of America). The security
benefits provided by the S*BGP proposals tend not to kick in until
they have reached almost global deployment. Worse yet, global
deployment of S*BGP requires each autonomous system to decide that it
has sufficient local incentive to undertake a potentially costly
upgrade of its network. As a result, the conventional wisdom argues
that global S*BGP deployment is infeasible.
This talk takes a more nuanced view. I argue that a
carefully-designed deployment strategy can create sufficient market
pressure to transition the majority of the Internet to S*BGP. I start
with an overview of the vulnerabilities of the Internet’s routing
system, and show how the S*BGP protocols can remedy these
vulnerabilities. I then present various models of incentives in
Internet routing, and finally propose a strategy that harnesses
autonomous systems local business objectives to drive global S*BGP
deployment. Our results use both theoretical arguments and
simulations to challenge the conventional wisdom by showing conditions
under which widespread deployment of S*BGP is possible.
Based on joint work with Phillipa Gill (Princeton), Pete Hummon
(AT&T), Zhenming Liu (Harvard), Jen Rexford (Princeton), and Michael
No background in networking or security will be assumed.
More information about the Busec