Project # 70
Trade-Offs Between Success-Attack Rate and Bandwidth Overhead for Timing-Based Webpage Fingerprinting Attacks

Team Lead:	Dong Quan Tran
Other Team Members:
Faculty Advisor(s):	Dr. Rémi A. Chou
Department:	CSE
Project Type:	REU
Title:	Trade-Offs Between Success-Attack Rate and Bandwidth Overhead for Timing-Based Webpage Fingerprinting Attacks
Abstract:	While low-latency anonymity systems like Tor encrypt data content, they remain vulnerable to traffic analysis. Specifically, Website Fingerprinting (WF) allows attackers to infer a user’s destination by analyzing packet timing, direction, and volume. Defenses like Tamaraw mitigate this by reshaping traffic into uniform patterns through fixed-interval transmissions and size padding. While Tamaraw effectively lowers attack-success percentage (ASP) in multi-webpage open-world settings, it is less robust in single-page settings, where ASP can exceed 90 percent. Although subsequent enhancements introduced more randomness and heavier padding to reduce ASP, they incurred substantial Bandwidth Overhead (BO). This project investigates the trade-off between ASP and BO within the Tamaraw framework. By adapting the algorithms from and utilizing contemporary datasets, we systematically sweep parameters—such as length bucket size and random padding strength—to quantify how much overhead is required to effectively suppress ASP in single-webpage evaluations.
Poster:	View Poster #70 (PDF)
Judges Scoring:	Submit Evaluation (judges only)

Project # 70Trade-Offs Between Success-Attack Rate and Bandwidth Overhead for Timing-Based Webpage Fingerprinting Attacks