High Precision Open-World Website Fingerprinting

Explore a comprehensive analysis of high precision open-world website fingerprinting in this IEEE conference talk. Delve into the effectiveness of traffic analysis attacks for identifying web pages browsed by clients using only packet metadata, particularly in the context of privacy technologies like Tor. Examine the limitations of previous website fingerprinting (WF) attacks in real-world scenarios with low base rates, and discover a new metric called r-precision that incorporates base rate into precision measurements. Learn about three novel classes of precision optimizers that significantly improve classifier performance, achieving a precision increase of more than 6 times for r = 1000. Investigate the scalability of WF classifiers to any open world set size and explore practical applications of precise classifiers in website fingerprinting, including identifying different types of websites, detecting sensitive clients, and overcoming WF defenses.