Hacker News Reader

Comment by doubleorseven | Hacker News Reader

"The entry WireGuard server will be able to see your source IP and which exit server the traffic is headed for, but it can’t see any of the traffic."

So server2 terminates the request twice? One for server1 and another time for the client who generated the request? I don't understand how it's possible for server1 to not be exposed to the data.

View on HN

You probably missed

> It’s a WireGuard tunnel being sent inside another WireGuard tunnel

Edit: replaced with a better diagram (and again, now based on example in [0]):

                   ▼    ▼                    ▼    ▼
                  YOU->NL1 tunnel           SE4->NL1 tunnel           PLAIN/TLS

            YOU ────────────────────► SE4 ───────────────────► NL1 ───────────────► CATPICS.COM

    On the wire:  YOU->SE4 traffic          SE4->NL1 traffic       NL1->CATPICS.COM traffic
                 ┌────────────────┐        ┌────────────────┐            ┌──────┐
    Inside:      │YOU->NL1 traffic│        │YOU->NL1 traffic│            │ DATA │
                 └────────────────┘        └────────────────┘            └──────┘

[0] https://mullvad.net/en/help/wireguard-and-mullvad-vpn/

This isn't how it works. If you actually pull down one of their multihop configurations - you'll see:

- the WireGuard public key for server 2

- the IP address for server 1

- a unique port for server2 on server 1

So all they're doing is a standard iptables redirect to the second host (which may or may not itself be under a WireGuard tunnel).

Well, I stand corrected, because I relied on their promo description. *shrug_emoji*

I replaced the diagram in the previous comment, take a look.

This can be very confusing indeed, since there are multiple ways of doing multihop. Please see my description in https://news.ycombinator.com/item?id=31012071.

The guide at https://mullvad.net/en/help/wireguard-and-mullvad-vpn/ only talks about how the config files does it. Which is completely different from how the app does it!

Your description is correct for the configuration files, yes! But it's not correct for the app. There are multiple ways of doing multihop with Mullvad. The config files use a simple redirect where each server has a unique port it's reachable over on all other servers. That's what the config files are doing.

But the app actually has a wg tunnel inside another wg tunnel. If you (on Linux) run `wg` (as root) in a terminal when it's connected with multihop you will see that it has two peers set up for the `wg-mullvad` interface, one peer is routed through the other.

So the only thing that SE4 can see is encrypted WireGuard traffic headed for NL1.

Reply on news.ycombinator.com