skip to main content
Department of Computer Science University of Colorado Boulder
cu: home | engineering | mycuinfo | about | cu a-z | search cu | contact cu cs: about | calendar | directory | catalog | schedules | mobile | contact cs
home · events · thesis defenses · 2010-2011 · 

Thesis Defense - Bauer

ECOT 831

Improving Security and Performance in Low Latency Anonymous Networks
Computer Science PhD Candidate

Tor is a volunteer-operated overlay network that provides low latency anonymous communication services to an estimated 200,000 daily users. While conventional wisdom dictates that the degree of anonymity provided by systems such as Tor increases as the user base grows, the most significant obstacles to Tor adoption today are its poor security and slow performance. In an effort to foster increased adoption, this thesis aims to understand and improve security and performance in low latency anonymous communication systems.

First, key findings are presented from a measurement study of anonymous Tor traffic, focusing on the application protocol distribution, geopolitical client and router distributions, and performance. In part, observations show that the vast majority of anonymous traffic by connection comes from interactive web browsing, but peer-to-peer file sharing protocols consume a disproportionately large amount of Tor's scarce bandwidth. This produces congestion that ultimately results in unfair and excessive delays for the majority of Tor users.

To reduce this congestion and offer a higher quality of service to interactive users, improved congestion and flow control algorithms inspired by techniques from IP and ATM networks are presented. A series of whole-network experiments indicate that delay-sensitive web users experience significantly faster web page responses and load times with the proposed congestion and flow control algorithms relative to Tor's current design.

New attacks on Tor users' anonymity are also presented. Experiments indicate that Tor is highly vulnerable to end-to-end traffic correlation attacks launched by weak adversaries who inflate their perceived bandwidth capacities to attract traffic. To reduce the resources needed to launch these attacks, a novel low-cost traffic analysis technique is developed that links clients with their respective destinations before any data traffic is sent. Defenses are proposed, many of which have been deployed on the live Tor network.

Committee: Dirk Grunwald, Professor (Co-Chair)
Douglas Sicker, Associate Professor (Co-Chair)
Shivakant Mishra, Associate Professor
Nikita Borisov, University of Illinois at Urbana-Champaign
Stefan Savage, University of California, San Diego

See also:
Department of Computer Science
College of Engineering and Applied Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
Send email to

Engineering Center Office Tower
ECOT 717
FAX +1-303-492-2844
XHTML 1.0/CSS2 ©2012 Regents of the University of Colorado
Privacy · Legal · Trademarks
May 5, 2012 (13:40)