Centrala begrepp
PURL is a machine-learning approach that effectively detects and sanitizes link decorations used for tracking, outperforming existing countermeasures in accuracy and reducing website breakage.
Sammanfattning
PURL introduces a novel approach to detect and sanitize link decorations used for tracking on websites. It leverages a graph representation of webpage execution to accurately identify ATS link decorations. The evaluation shows PURL's superior performance in accuracy, precision, and recall compared to existing countermeasures. By analyzing the prevalence of link decorations across top-million websites, PURL detects widespread abuse of link decoration for tracking purposes. The key contributions include proposing an automated machine learning approach, measuring the prevalence of link decorations on websites, and generating a filter list for privacy-focused browsers.
Statistik
PURL achieves 98.74% accuracy in detecting ATS link decorations.
PURL significantly outperforms existing countermeasures by at least 7.71% in precision and 6.43% in accuracy.
PURL detects that 73.02% of sites abuse link decorations for tracking.