When a fingerprinting model is trained on traffic collected from one geographic vantage point and tested on traffic from a different continent, the HTTPS-only open-world FPR at 0.5 recall increased by factors ranging from 2.8x (EU-West-2) to 50.3x (Africa) relative to the same-vantage baseline — despite 60-way closed-world accuracy remaining above 0.99 across all vantage-point pairs (Table 5). For Tor traffic the effect was weaker but still reached 25.2x (Asia-Pacific Southeast-1), showing path diversity also disrupts Tor-based fingerprinting.
From 2025-walsh-improved-open-world-fingerprinting — Improved Open-World Fingerprinting Increases Threat to Streaming Video Privacy but Realistic Scenarios Remain Difficult
· §5.1, Table 5
· 2025
· PoPETs 2025
Implications
Geographic relay diversity is a first-order defense: training a fingerprinting model at one vantage point and deploying it globally degrades FPR by up to 50x, so distributing circumvention relay infrastructure across continents inherently defeats centrally-trained classifiers.
Never rely on closed-world accuracy to evaluate fingerprinting resistance — a less than 1% accuracy drop masks order-of-magnitude FPR degradation in the open-world; large open-world evaluations from geographically diverse vantage points are required.