cross-posted from: https://lemmy.world/post/3754933
While experimenting with ProtonVPN’s Wireguard configs, I realized that my real IPv6 address was leaking while IPv4 was correctly going through the tunnel. How do I prevent this from happening?
I’ve already tried adding
::/0
to theAllowedIPs
option and IPv6 is listed as disabled in the NetworkManager profile.
The bad practise would be to entirely disable IPv6. #ShittySysAdmin
Broke: disabling ipv6 because it’s causing security holes with software not originally built for it
Woke: disabling ipv6 because hex ip strings are too hard to type
Honest to Tux, I’ve never memorized a single IPv6 since its public release; ever.
Well I memorized mine: ::1
Me either. I have a text file on the desktop that I use to copy paste the loopback address on the rare rare times I actually need it lul
The whole Silicon Valley, governments and economic infrastructure of the world hang by a thread made of bootleg shortcuts.
Haha, yeah that’s what ProtonVPN support recommended to me and I am not sure if I want to disable IPv6 globally like that. I feel (with my inexperienced brain) that there’s gotta be a way to do it on a per connection basis.
Maybe try forcing only IPv4 connections for the traffic you want to keep private?
Your biggest leak maybe IPv6 DNS which is probably not that big of a deal really…
In this case, disabling IPv6 is actually the right move. If the VPN provider doesn’t support IPv6, then there’s no way to allow to allow IPv6 Internet traffic without causing a leak/VPN bypass. If you block IPv6 via firewall or routing it to a dead-end, it will add delays as things try IPv6, timeout, and fall back to IPv4. If you just remove the IPv6 address from the Internet interface, you have to also make sure it doesn’t get re-added by SLAAC/DHCPv6 or other interface changes (switching wifi networks, etc). As dumb as it seems, disabling IPv6 or switching to a provider that supports it are probably the best options.
In this case, disabling IPv6 is actually the right move. If the VPN provider doesn’t support IPv6, then there’s no way to allow to allow IPv6 Internet traffic without causing a leak/VPN bypass.
The right move for the VPN provider is to support IPv6. The right move for the user is to take their business to a provider that does.
Yes, I agree - using a VPN that supports IPv6 is preferred. If they don’t want to/can’t change providers, then disabling IPv6 is the best way to prevent IPv6 leaks without negatively affecting network performance.
I do the same, but would like to know the correct way to hide ipv6
Unironically that’s one thing we do to anonymize traffic at work. We could do NAT via IPV6 but what’s the point when IPV4 already helps mask who is doing what in the office.
IPv6 is broken by design /s.
IPv4 and IPv6 are two different network stacks. Your IPv4 stack is hidden behind wireguard, but not the IPv6 one.
The correct way to fix your issue is to setup a second witeguard tunnel for IPv6, and route IPv6 traffic through it.
Edit: many comments advise to block outbound IPv6 traffic. Don’t do that! It will add latency to all your requests as you will have to wait for them to timeout.
Unfortunately I can’t change the tunnel as it is provided by ProtonVPN.
Apparently they advise to disable IPv6. I’m a bit disappointed 😞
If you disable IPv6 at the kernel level there’s no extra latency as nothing even tries to connect to an IPv6 address. It’s a shame to have to do it, but does fix the issue.
Disabling it is fine indeed, but I saw many comments advising to block outbound traffic, so I warned against that.
Either fix IPv6 in the Wireguard tunnel or, if the tunnel isn’t under your control, set up your firewall to drop all IPv6 traffic directed towards the wider internet. You could also gimp your normal network interface by disabling IPv6 on it if you always have the VPN on anyway.
Disabling IPv6 system wide will cause all kinds of fun and interesting random failures in tons of applications. You can drop all the routes you want, but I wouldn’t remove the link local addresses or the ::1 address unless you like debugging failed assertions in your system logs.
For me, it was rather the opposite: when dropping IPv6 packets, applications would often hang and behave weirdly. Disabling IPv6 completely would mean they’d stop trying to do anything on IPv6 and function well.
Dropping all IPv6 packets would cause problems but if you only drop the packets destined for the internet you should be fine, I believe. You may also need to configure your DNS to prefer A records over AAAA, though, but if your VPN only supports IPv4 I assume your VPN’s DNS server doesn’t accept AAAA queries.
My ISP doesn’t even support IPv6
My country doesn’t really support IPv6.
Vodafone gang!
Time to get a better ISP
Wireguard easily supports dual stack configuration on a single interface, but the VPN server must also have IPv6 enabled. I use AirVPN and I get both IPv6 and IPv4 with a single wireguard tunnel. In addition to the ::/0 route you also need a static IPv6 address for the wireguard interface. This address must be provided to you by ProtonVPN.
If that’s not possible, the only solution is to entirely disable IPv6.
Yes, agree with you, especially your last comment: disable entirely IPv6
The true enterprise solution is to disable IPv6, if you disable IPv4 you might be surprised to see how many sites and services, even those run by large companies require IPv4.
I just disable ipv6 on my router to simply my firewall and such… Won’t use it until necessary
This is a real backwards approach to this.
Why? I don’t miss anything
You’re missing a chance to help cool tech moving forward :)
Using networking namespaces may be an option https://gist.github.com/zenofile/d3ada8a1becb8c88aadffa68cdb18350
You can either route IPv6 through the tunnel, or if the tunnel doesn’t support it, you have to block it somehow if you don’t want it going out to the Internet. You can do that by creating a firewall rule to block it on your host or network, or by disabling IPv6 entirely at the host or network level.
You can either route IPv6 through the tunnel
That’s what I am trying to do with the
::/0
option but it doesn’t seem to work. I am guessing because the tunnel itself doesn’t have IPv6?Add a dummy IPv6 on the WireGuard interface, like a completely random fd00::/128 address on it so it thinks it’s IPv6 enabled. It’ll then just go nowhere as the remote end won’t accept it. You can then drop it at the firewall level before it goes into the tunnel to save some bandwidth.
Thanks for the suggestion! It gave me the idea to try setting the connection profile’s IPv6 setting to either
Link Local
orIgnored
. Both those options, along with the::/0
inAllowedIPs
, seem to work according to ipleak.net, but I am unsure what the bigger ramifications of this change are.It’s going to still send that IPv6 traffic through the tunnel and get discarded at the other end but that’s about it. You can firewall that off if you want, ideally with a reject rule so things don’t have to timeout before realizing it goes nowhere.
Just disable IPv6, if you don’t wanna, try a configuration tool like PiVPN, which can setup wireguard automatically
I use a magisk module to disable ipv6
decreasing security for a… privacy(?) increase 🤔
We’re not talking about privacy in that context here
We’re trying to patch a leaking ipv6 which I gave my solution to.
Don’t use it if you don’t like it lmao
I suppose but my point was that rooting your device decreases security immensely via crippling the android security model.
That is true… but it’s always about your threat model
Here, the ipv6 leaks are the priority