IPv6 Leak: Test, Fix and Decide Whether to Disable IPv6
An IPv6 leak means IPv6 traffic bypasses your VPN and reveals your real network. Run the IPv6 leak test, fix it in the VPN, then decide whether to disable IPv6.
In this guide
An IPv6 leak happens when your VPN carries only IPv4 while your connection also has IPv6: the operating system prefers IPv6, so sites that support it see your real IPv6 address and ISP instead of the VPN. Test with the VPN on; fix it in the VPN’s settings first, and disable IPv6 on the device only as a fallback.
Test it now
CheckingRunning against our own test servers. Nothing is stored.
Open the IPv6 testWhat is an IPv6 leak?
An IPv6 leak is IPv6 traffic that leaves your device outside the VPN tunnel while the VPN is connected. IPv6 is the current Internet Protocol, defined in RFC 8200 as the successor to IPv4 with addresses widened from 32 to 128 bits, and many home connections now run both side by side (dual stack). A VPN built for IPv4 routes the IPv4 address space into its tunnel and says nothing about IPv6: the IPv6 address your router handed out stays active, the IPv6 route to your ISP stays in place, and nothing stops a packet from using it.
The operating system then makes the leak the default. RFC 6724, the standard for choosing between addresses, ships a policy table that prefers native IPv6 destinations over IPv4 (precedence 40 for ::/0 against 35 for IPv4-mapped addresses), and that order is the order in which an application tries addresses. For any site that publishes an IPv6 address, your browser tries IPv6 first, the connection goes straight out through your ISP, and the site logs an address that belongs to your home network, your city and your provider. IPv4 traffic still travels through the VPN, which is why the leak is easy to miss: the app says connected, and an IPv4-only check page agrees. A resolver reached over IPv6 bypasses the tunnel the same way, so a DNS leak and an IPv6 leak often appear together.
Test for an IPv6 leak in 30 seconds
The test compares what the VPN should show with what each protocol actually shows. Connect the VPN first, then:
- Open the IPv6 test. It fetches from an IPv4-only, an IPv6-only and a dual-stack hostname and reports a verdict (IPv4 only, dual-stack with IPv6 or IPv4 preferred, IPv6 only, or offline), the IPv6 address it saw, and whether IPv4 and IPv6 come from the same network. With a VPN connected, the clean results are IPv4 only (the VPN blocks IPv6) or dual-stack with both addresses in the VPN’s network. An IPv6 address from a different network than the IPv4 address is the leak.
- Confirm the owner with the ASN lookup. If the IPv6 address resolves to your ISP’s ASN while the IPv4 address resolves to the VPN provider, IPv6 is leaving through your ISP.
- Run the VPN test, which runs the IP, DNS, WebRTC and IPv6 leak checks together with a pass or fail per test and the fix for each.
- Run the DNS leak test. It lists the resolvers that answered queries for the site’s own test zone, with each resolver’s ASN, location and any EDNS Client Subnet it forwarded. A resolver with an IPv6 address means DNS itself travels over IPv6; a resolver owned by your ISP means DNS is leaving the tunnel.
test-ipv6.com uses the same method: client-side JavaScript fetches from hostnames that force IPv6-only or dual-stack behavior, and pass/fail plus timing decides the result. That is also why a leak test must run on each device separately; the result describes the device’s own address selection, not the network.
Why it happens: no IPv6 in the VPN app, split tunneling, kill-switch gaps, router advertisements, reconnects
No IPv6 support in the VPN app. The tunnel carries IPv4 only and the app installs no IPv6 route and no IPv6 firewall rule. Vendor documentation makes the gap visible: one provider’s support page lists the platforms on which its apps tunnel IPv6 and says that on some platforms the app instead turns IPv6 off so the address cannot be exposed, which is the leak protection described below.
Manual configurations. A WireGuard peer’s AllowedIPs is, in WireGuard’s description, a cryptokey routing table: a packet goes to the peer whose allowed IPs match its destination, and 0.0.0.0/0 is a wildcard for any IPv4 destination. A configuration with 0.0.0.0/0 but without ::/0 routes every IPv4 packet into the tunnel and no IPv6 packet at all.
Split tunneling and kill-switch gaps. An app or site excluded from the tunnel is excluded on both protocols, DNS included; a kill switch that blocks IPv4 outside the tunnel but not IPv6 lets IPv6 through during reconnects and while the tunnel is down.
Router advertisements and reconnects. IPv6 addresses arrive from the router by autoconfiguration, without a lease exchange, so a new network or a reconnect can add an IPv6 address after the VPN finished its setup; a laptop that moves from home to a cafe keeps the VPN session and gains a fresh IPv6 path.
Fix 1: Turn on the VPN’s IPv6 leak protection or pick a VPN that tunnels IPv6
The correct fix is in the VPN, because it keeps IPv6 working for everything else. There are two designs:
- IPv6 inside the tunnel. The app assigns an IPv6 address from the VPN’s range and routes ::/0 through the tunnel as well as 0.0.0.0/0. The IPv6 test then shows an IPv6 address in the VPN’s network, and the ASN lookup confirms it.
- IPv6 blocked. The app adds a firewall rule that drops all IPv6 outside the tunnel, often labeled IPv6 leak protection. The IPv6 test then reports IPv4 only while the VPN is connected, and nothing on the device can use IPv6 until you disconnect.
Find the IPv6 switch or leak protection option in the app’s settings, turn it on, reconnect, and repeat the test. For a manual WireGuard or OpenVPN setup, add ::/0 to the routes the tunnel captures or block IPv6 on the physical interface with the OS firewall. If the app offers neither and its documentation never mentions IPv6, the app does not handle it, and the decision in the next two sections applies. Tunneled IPv6 keeps dual-stack sites and services working; blocking is simpler, with nothing to route.
Fix 2: Disable IPv6 on the device (Windows, macOS, Linux, Android, iOS, router) and what you lose
Disabling IPv6 removes the path the leak uses, on every network, for every app, and removes IPv6 for everything that benefits from it, so treat it as the fallback. Steps are from the vendors’ current documentation.
| Platform | Where | What it does | Vendor note |
|---|---|---|---|
| Windows 10/11, one adapter | Adapter properties, clear Internet Protocol Version 6 (TCP/IPv6); or Disable-NetAdapterBinding -Name "AdapterName" -ComponentID ms_tcpip6 in PowerShell | Unbinds IPv6 from that adapter | Microsoft: unbinding “can result in an unsupported Windows configuration” |
| Windows, system-wide | Registry key HKLM\SYSTEM\CurrentControlSet\Services\Tcpip6\Parameters, DWORD DisabledComponents: 0x20 prefers IPv4, 0xFF disables; restart required | 0x20 keeps IPv6 bound but tried second; 0xFF disables it on all interfaces | Microsoft recommends 0x20 “instead of disabling IPv6”; 0xffffffff is a documented mistake that delays startup |
| macOS | System Settings > Network > your connection > Details > TCP/IP > Configure IPv6 | Automatically is the default; Link-local only keeps IPv6 off the internet | Apple documents the menu; “Link-local only” is Apple’s wording for limiting IPv6 to the local network |
| Linux | sysctl -w net.ipv6.conf.all.disable_ipv6=1, persisted in sysctl.conf | Sets every interface’s disable_ipv6; removes the interface’s IPv6 addresses and routes | Kernel docs: default is FALSE (IPv6 enabled) |
| Android, iOS | No system switch | If the router hands out IPv6, the phone uses it | Fix it in the VPN app or at the router |
| Router | IPv6 or WAN IPv6 page (NETGEAR: ADVANCED > Advanced Setup > IPv6 > Internet Connection Type) | Stops the router advertising IPv6 to every device | Labels vary by vendor; this turns IPv6 off for the whole home |
Microsoft’s guidance is direct about the cost: “We don’t recommend that you disable IPv6 or IPv6 components or unbind IPv6 from interfaces. If you do, some Windows components might not function,” because “Windows is tested with, and some products and features expect, IPv6 to be bound and functional.” Even 0xFF cannot completely disable IPv6, since the system uses it internally. The recommended 0x20 makes Windows prefer IPv4, so dual-stack sites are reached over IPv4 through the tunnel while IPv6 stays available; IPv6-only destinations can still leak, so re-test after setting it.
What you lose on any platform: direct public addresses for every device, which matters behind carrier-grade NAT where the IPv4 address is shared and port forwarding is impossible; peer-to-peer connections that would otherwise skip NAT; and the IPv6-only services the IPv6 test scores readiness for. The IPv4 vs IPv6 guide sets out the differences.
Is it good to disable IPv6? Microsoft’s guidance, IPv6-only networks, CGNAT and when disabling is justified
Disabling IPv6 is justified in one situation: the VPN you depend on cannot tunnel or block IPv6, you cannot change it, and the privacy of your real address matters more than what IPv6 provides. Everything else argues for leaving it on.
Microsoft’s position, quoted above, is the reference for Windows, where IPv6 has been a mandatory component since Windows Vista. Sony’s PS5 setup page says the console supports IPv6 connections, though not IPv6-only networks. Behind CGNAT, IPv6 is the only way an inbound connection can reach a device in your home directly, because the shared IPv4 address is not yours to forward ports on. And an IPv6-only network, which some mobile carriers run with NAT64 and DNS64, has no IPv4 to fall back to: a device with IPv6 disabled there is offline, which the IPv6 test reports when its IPv4-only fetch fails.
If the reason is a VPN leak, fix the VPN. If the reason is a broken IPv6 path (sites hang, then load), prefer IPv4 with Microsoft’s 0x20 value or report the path to your provider rather than removing IPv6 entirely.
Does disabling IPv6 make internet faster? What the evidence supports
The evidence does not support disabling IPv6 for speed. Modern clients implement Happy Eyeballs version 2 (RFC 8305): when a hostname has both kinds of address, the client starts the preferred family first and the other after a short delay, with a recommended default of 250 milliseconds, and keeps whichever connects first. The RFC exists because “specific addresses or address families (IPv4 or IPv6) may be blocked, broken, or sub-optimal on a network,” and its answer is to race them, not to turn one off, so a bad IPv6 path costs a fraction of a second per new connection and a good one costs nothing. Older selection without the race could stall much longer; test-ipv6.com notes that a broken IPv6 route under plain RFC 3484 selection could take “towards 75+ seconds” to give up, the experience that gave the advice its reputation.
Measure instead of assuming. The IPv6 test’s dual-stack row shows which family won the race on your connection; IPv4 winning where IPv6 works is a sign of a slow IPv6 path worth reporting to the provider. The speed test run before and after any change shows whether the change did anything on your line.
IPv6 privacy extensions (temporary addresses) vs leaks
A leak reveals your network; privacy extensions change which address inside that network you show. RFC 8981 defines temporary addresses: the device generates a random interface identifier, uses it for outgoing connections, and replaces it on a schedule, by default a preferred lifetime of one day and a valid lifetime of two. RFC 6724 ranks a temporary source address above a public one (rule 7), which is why the address the IPv6 test shows is normally the temporary one and differs from the stable address in your settings. On Linux the kernel’s use_tempaddr sysctl controls it (0 disables privacy extensions, 1 enables them but prefers public addresses, values above 1 prefer temporary addresses), with a kernel default of 0 for most devices.
The limit is in the standard itself: RFC 8981 notes that the network prefix “also serves as a constant identifier,” so a changing suffix still points at the same home network and provider, and an on-link observer such as the router sees every address. Temporary addresses stop long-term tracking of one device by its address; they do nothing about a leak, because the leaked address, temporary or not, carries your prefix.
Re-test after every reconnect and network change
IPv6 configuration is dynamic. A router advertisement can add an address after the VPN connected, a reconnect can bring the tunnel up before the IPv6 route is blocked, and an update can change a default. Test at these moments:
- After connecting the VPN on a new network (hotel, office, cafe, mobile data).
- After the VPN reconnects by itself, which the app usually logs.
- After updating the OS, the VPN app or the router firmware.
- After changing any IPv6 setting, including the router’s.
Run the VPN test, which covers IPv6 alongside IP, DNS and WebRTC; a browser that exposes a real address through WebRTC undoes a perfect tunnel just as surely. On phones and tablets, run it in the device’s own browser; the result applies only to the device that runs it.
Sources
- RFC 8200: Internet Protocol, Version 6 (IPv6) Specification
- RFC 8981: Temporary Address Extensions for Stateless Address Autoconfiguration in IPv6
- RFC 6724: Default Address Selection for Internet Protocol Version 6 (IPv6)
- RFC 8305: Happy Eyeballs Version 2: Better Connectivity Using Concurrency
- Microsoft Learn: Guidance for configuring IPv6 in Windows for advanced users
- Linux kernel documentation: IP Sysctl (disable_ipv6, use_tempaddr)
- Apple macOS User Guide: Use DHCP or a manual IP address on Mac (TCP/IP settings)
- Apple AirPort Utility User Guide: Modify IPv6 settings (definition of Link-local only)
- Proton VPN Support: Prevent IPv6 VPN leaks (vendor documentation, cited not endorsed)
- WireGuard: Conceptual overview (Cryptokey Routing)
- test-ipv6.com: FAQ and methodology
- NETGEAR Support: How do I set up an IPv6 Internet connection on my NETGEAR router?
- PlayStation Support: How to set up an internet connection on PlayStation consoles
Questions
Is IPv6 recommended?
Yes, for normal use. Microsoft does not recommend disabling IPv6 and says some Windows components might not function without it; consoles such as the PS5 support it; and on ISPs that share IPv4 through CGNAT, IPv6 is the only path that gives your home its own public addresses. Disable it only to close a leak you cannot fix otherwise.
What should I do if my IPv6 address isn't being detected?
With a VPN connected, that is usually the VPN blocking IPv6 on purpose, which is the safe state. Without a VPN, check in this order: the provider delivers IPv6 on your line, the router’s IPv6 or WAN IPv6 setting is on automatic, and the device has IPv6 enabled. Reconnect and re-run the IPv6 test after each change.
How do I check if my DNS is leaking?
Run the DNS leak test with the VPN connected. It shows which resolvers answered the test queries, with their owner and location. A resolver belonging to your ISP, or a public resolver reporting your real network prefix through EDNS Client Subnet, means DNS leaves the tunnel. Resolvers with IPv6 addresses show whether DNS itself travels over IPv6.
Does disabling IPv6 make internet faster?
Not in general. Modern clients race IPv4 and IPv6 connections (Happy Eyeballs, RFC 8305) and keep whichever answers first, so a slow IPv6 path costs milliseconds, not seconds. If the IPv6 test shows IPv4 winning the dual-stack race, the IPv6 path is worth reporting to your provider; measure before and after with the speed test rather than assuming.
Back to your own address: what is my IP address

