Using tornettools-based simulations with historical Tor consensus data scaled to 1% of the real network (80 relays), the adversary's exit-observation probability p̂exit grows monotonically with adversary-controlled bandwidth: a single exit relay at 148 Mbps yields p̂exit ≈ 2.13%, and 5 adversary-controlled exit relays push p̂exit above 10%.
From 2026-fan-activeflowmark-assessing-tor — ActiveFlowMark: Assessing Tor Anonymity under Active Bandwidth Watermarking
· §V.B, §VI.A
· 2026
· arXiv preprint
Implications
Designs that steer circuits away from high-bandwidth exit relays (e.g., AS-aware path selection, Counter-RAPTOR) directly reduce the exit-observation component of the watermarking attack surface.
Bridge or pluggable-transport configurations that never expose an adversary-controlled Tor exit relay eliminate the Sniffer vantage point required for NATA; deployments should consider end-to-end tunnel designs that bypass the public Tor exit layer.