WireGuard for Windows Hits v1.0 Stability Milestone

Researcher tested Mullvad VPN's WireGuard exit-IP allocation and found it's surprisingly deanonymizing. By rotating public keys and querying nine Mullvad servers, the author mapped exit-IP pools and observed only 284 distinct exit-IP combinations across 3,650 keys — far fewer than the trillions of theoretical combinations. The pattern: each assigned IP sits at nearly the same percentile within its server pool, implying Mullvad deterministically maps a key to a proportional index in each pool. Correlated indexes across servers with identical pool sizes suggest a seeded RNG or deterministic function is used to pick indexes, not per-connection randomization. This makes exit IPs a stable fingerprinting vector tied to rotating WireGuard keys, with privacy implications for users and VPN design.

Researcher testing Mullvad’s WireGuard exit IP allocation found the provider deterministically maps a user’s public key to a fixed exit IP index per server, rather than assigning a freshly randomized IP on each connection. By cycling pubkeys against nine servers and collecting 3,650 datapoints, the author mapped each server’s exit-IP pools and discovered only 284 distinct cross-server combinations — far fewer than the combinatorial space — because Mullvad appears to pick corresponding percentile positions within each server’s IP pool using a seeded RNG. That behavior makes exit-IP sets a stable fingerprinting vector tied to rotating keys (or persistent when keys don’t rotate), raising deanonymization risks for users. The finding matters for VPN privacy design and threat models.