Civilian reference receivers on NATO’s eastern edge recorded something worse than noisy GPS last year: whole stretches of the sky where fixes simply stopped. A study of about 10 TB of ground-station data from Estonia, Latvia and Lithuania traces how Russian interference went from an irritant to a standing denial, and Estonia took the hardest hit.
Johan Juul, of Tallinn University of Technology, laid out the results at ION GNSS+ 2026. His team ran an LSTM autoencoder over CORS measurements logged between 2023 and 2025, testing whether fixed antennas can document what aircrews and shipmasters describe only in complaints.
The method worked. Baseline signal strength at NJO1, a site hard against the Russian border, had run a healthy 48-50 dB-Hz; it slid to 30-35 dB-Hz. Over 2025 that single receiver counted 51 full outages and 130 depression events. Nearby TOIL notched 97 outages and then stopped transmitting altogether.
Timing is the striking part. When Ukrainian drones struck the Ust-Luga gas complex in August 2025, NJO1 lost 97.7 percent of satellite reception for 72 hours, both L1 and L2 swamped. Nine hours of simultaneous blackout hit both receivers when a drone reached Auvere in March 2026. To Juul, the March picture points to three possibilities: stronger transmitters, a shifted geometry, or several emitters active together.
Lithuania fared differently. LITPOS sites held 43-44 dB-Hz and saw outages lasting hours, plausibly because they sit farther from the emitters with less clear sightlines. The five Lithuanian sites emerged from the 2025 Zapad war games, an exercise built around electronic attack, without meaningful degradation.
Anyone still filing interference under routine annoyance should think again, because fixed antennas document denial that flight and maritime reports miss, and the exploitation of a weak user signal looks deliberate.
