Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Yggdrasil – Early-stage implementation of an end-to-end encrypted IPv6 network

github.com

Post reply on HN
Print view View original

81–90 of 109 posts

Re: Yggdrasil – Early-stage implementation of an end-to-end encrypted IPv6 network

#81

Post by infogulch »

neilalexander wrote: post #10

We're prepping for a major new release too — information here: https://yggdrasil-network.github.io/2021/06/19/preparing-for...

https://yggdrasil-network.github.io/2021/06/19/preparing-for...

> the root is the node with the lowest ed25519 public key, rather than the highest sha512sum hash of the public key

With this scheme, could a bad actor decide to choose a poor key just to be the neighbor of a target in (edit) keyspace? Ordering by the hash of a public key means that the order is protected by the hash function's preimage resistance; does the generation of a ed25519 key have a similar protection?

Re: Yggdrasil – Early-stage implementation of an end-to-end encrypted IPv6 network

#82

Post by miloignis »

infogulch wrote: post #81
neilalexander wrote: post #10

We're prepping for a major new release too — information here: https://yggdrasil-network.github.io/2021/06/19/preparing-for...

https://yggdrasil-network.github.io/2021/06/19/preparing-for... > the root is the node with the lowest ed25519 public key, rather than the highest sha512sum hash of the public key With this scheme, could a bad actor decide to choose a poor key just to be the neighbor of a target in (edit) keyspace? Ordering by the hash of a public key means that the order is protected by the hash function's preimage resistance; does…

Even with preimage resistance, it seems like it wouldn't be insane to just burn cycles until you got one close enough, bitcoin-style.

Dealing with attackers in a system like this seems very challenging, though very worthwhile in the end! Maybe something web-of-trust-y...

Re: Yggdrasil – Early-stage implementation of an end-to-end encrypted IPv6 network

#83

Post by infogulch »

How does Yggdrasil compare to Wireguard? A github search shows that yggrasil-go uses wireguard-tun project as the tun driver; does it relate in any other way? The main problem/use case is different of course (Wireguard is a manually configured point-to-point vpn with e2ee, where yggdrasil is an internet-scale overlay network with e2ee.), but I mean at a low level, protocol, encryption, etc.

Re: Yggdrasil – Early-stage implementation of an end-to-end encrypted IPv6 network

#84

Post by infogulch »

miloignis wrote: post #82
infogulch wrote: post #81

Earlier quoted context omitted.

https://yggdrasil-network.github.io/2021/06/19/preparing-for... > the root is the node with the lowest ed25519 public key, rather than the highest sha512sum hash of the public key With this scheme, could a bad actor decide to choose a poor key just to be the neighbor of a target in (edit) keyspace? Ordering by the hash of a public key means that the order is protected by the hash function's preimage resistance; does…

Even with preimage resistance, it seems like it wouldn't be insane to just burn cycles until you got one close enough, bitcoin-style. Dealing with attackers in a system like this seems very challenging, though very worthwhile in the end! Maybe something web-of-trust-y...

My point is that I don't think there's anything during key generation that requires the resulting key to be secure / chosen 'well', so an attacker might exploit key generation as a way to target a particular spot in the line, where having an insecure/easily compromised key doesn't matter to them even though that may be detrimental to the health of the network. But at least with preimage resistance any public key is just as likely as any other to get a particular spot that they desire so there's no incentive to exploit key generation to get it, they might as well generate a bunch of secure keys if they want to use a bitcoin-style brute force strategy.

Re: Yggdrasil – Early-stage implementation of an end-to-end encrypted IPv6 network

#85

Post by natural219 »

infogulch wrote: post #83

How does Yggdrasil compare to Wireguard? A github search shows that yggrasil-go uses wireguard-tun project as the tun driver; does it relate in any other way? The main problem/use case is different of course (Wireguard is a manually configured point-to-point vpn with e2ee, where yggdrasil is an internet-scale overlay network with e2ee.), but I mean at a low level, protocol, encryption, etc.

I’d also be curious about this. Say for the use case of running a small private chat server hosted on a home network; does either of these seem better suited, or are they just different architectures that can handle mostly similar things?

Re: Yggdrasil – Early-stage implementation of an end-to-end encrypted IPv6 network

#86

Post by woah »

neilalexander wrote: post #22
4bpp wrote: post #16

One thing that wasn't clear to me from the documentation: What's the typical latency you observe with this network? Does the routing take physical distance/observed delays into account in some way, or could you wind up with short (in network space) paths that in reality bounce a packet back and forth between the US and New Zealand repeatedly?

Latency is dependent on the underlying peering connections. Yggdrasil will try wherever possible to take the shortest paths in network space, but yes, it's possible those could be physically indirect if the underlying peerings are indirect. Generally on the public network (which is probably a couple thousand nodes in number, where people have contributed a number of public nodes and have interconnected them) we see v…

Are you not routing based on latency and reliability like i.e. Babel does? You're only routing based on number of hops?

Re: Yggdrasil – Early-stage implementation of an end-to-end encrypted IPv6 network

#87

Post by fragileone »

number6 wrote: post #27
neilalexander wrote: post #21

Earlier quoted context omitted.

Yggdrasil builds a multi-hop IPv6 overlay network using peer-to-peer connections. You can string a whole bunch of nodes together using whatever means (cables, wireless or TCP peerings over the internet or any other network) and you get a full-mesh network where everyone can reach everyone else. It's designed to be as minimal-configuration as possible and scalable in a way that many other mesh routing protocols aren't…

So it's a kind of VPN?

A mesh VPN, or kinda a mesh TOR really.

Re: Yggdrasil – Early-stage implementation of an end-to-end encrypted IPv6 network

#89

Post by boardwaalk »

aleken wrote: post #88

I have two devices split by VNET and not routed out to the internet. I connect those two and a VPS to create a small Yggdrasil network. This allows me to access all three devices from “anywhere”. Would use again.

What does this give you that Wireguard wouldn’t? (Honest question, I have no idea.)

Re: Yggdrasil – Early-stage implementation of an end-to-end encrypted IPv6 network

#90

Post by aleken »

boardwaalk wrote: post #89
aleken wrote: post #88

I have two devices split by VNET and not routed out to the internet. I connect those two and a VPS to create a small Yggdrasil network. This allows me to access all three devices from “anywhere”. Would use again.

What does this give you that Wireguard wouldn’t? (Honest question, I have no idea.)

Good question. I could achieve the same with Tailscale or Innernet, both using Wireguard. So perhaps my answer has to be that Yggdrasil tickles my interest in mesh networking.
Post reply on HN