PHYSICS DOESN'T
LIE.
A device can claim to be anywhere. It cannot claim to have traveled faster than physically possible between two real transactions.
// Geo-Velocity Scoring
Every location pair is scored by Haversine great-circle distance ÷ elapsed time. The result classifies straight into a risk tier — no manual thresholds per merchant.
// Fail-Closed Design
- Bad or missing input → CRITICAL, not "unknown".
- Low velocity alone isn't automatically clean.
- VPN/Tor/proxy context can enhance a low-velocity signal to HIGH.
GPS Says One Thing. IP Says Another.
Mock-GPS apps spoof the device's own location API convincingly. What they can't spoof is where the network traffic actually originates. NonaShield cross-references device-reported GPS against edge-derived IP geolocation — independently, on every request.
Mismatch ≥300km: LOW severity flag
Mismatch ≥1500km: HIGH severity — GPS vs IP geolocation diverge by more than a country.
Anonymous-IP databases (the class of data that flags "this IP is a known VPN/Tor exit") are notoriously incomplete across every vendor, ours included on this deployment. Rather than market a single silver-bullet VPN/Tor flag, NonaShield leans on ASN-organization heuristics and a maintained hostile-ASN list at the edge, combined with behavioral and velocity signals — so no single missed database entry becomes a single point of failure.
The Verdict.
"GPS is a claim. Velocity, IP geolocation, and a learned home zone are three independent witnesses."