Peer-to-peer DNS: Decentralizing the internet’s domain names

In light of the recent WikiLeaks domain name problems and COICA DNS hijacks, I’ve been thinking more and more about some kind of peer-to-peer DNS system. The peer-to-peer model is the same system used by BitTorrent, and is different from the centralized model used by most internet services. This would decentralize the internet’s domain names and make it more difficult for governments and corporations to manipulate the internet. A peer-to-peer system also has the advantage of different fault-tolerance characteristics. It could also improve the DNS response times of the internet, due to the fact that a peer might be right next to you and already have the DNS information your computer is looking for.

Right now, people are sending each other the IP addresses of sites like WikiLeaks (213.251.145.96) but this only works while free speech is allowed and normal DNS is still working. A system needs to be created to address this.

Both Windows and *nix-based operating systems have support for HOSTS files. A HOSTS file is just a human-readable list of domain/host names and IP addresses that the network stack looks through first before making a DNS query for a domain name. Usually, the HOSTS file on most computers is almost empty, containing only entries for localhost and such.

At the most basic level, you could download a text file via BitTorrent containing a list of DNS/IP address pairs, and copy that into your HOSTS file. This works, and acts as a temporary solution, but there are a few problems with this approach:

• It requires manual edits to the HOSTS file. Users with less technical experience may not understand how to do this.
• There is no way to verify the validity of the information – this is only established by word-of-mouth and peer count. Someone could plant fake IP addresses and phish passwords – or execute other attacks – and it could take a while before this is found out.
• There is no easy way to update the entries with new information, or combine multiple lists of DNS information. You would need to download a new file every time entries are changed, which would lead to a ton of files being seeded on BitTorrent trackers with people unsure of which ones are the most recent. You would probably need a script to remove duplicate entries, and you would need to merge all the changes manually to resolve any conflicts. The HOSTS file could also get quite large; there would be performance issues with searching through the file every time a DNS query is made.

It’s obvious that this isn’t very scalable beyond a few pages of entries. A better solution would be to set up an actual DNS server, either just for your own computer or your entire local network, and route DNS requests through there first. Storing the IP addresses as 4 or 16 bytes and using an optimized data structure for the domain names would be more efficient. New entries would be downloaded automatically via methods similar to peer exchange and DHT in BitTorrent.

• Protocol encryption would secure the exchange of entries between peers.
• Asymmetric cryptographic signatures can be used to prove that new DNS entries for a domain were written by the same entity as the old DNS entries.
  • This means that when someone forwards you a new IP address for a domain, you can tell that the new IP address was originally from the same person that wrote the old IP address.
• I’m not sure how you would prove or claim ownership of a domain name, though. This is one of the big problems. The owner(s) of a site should be the only ones who have the authority to change or set DNS entries.
• Hash collisions would be bad in this system – two DNS entries that share the same hash would produce undesirable behavior.
• What happens if two peers upload conflicting DNS entries at the same time? Some peers would end up seeing different IP addresses, and then when the two entries diffuse far enough to be seen simultaneously by a single peer, which one does that peer select?

As you can see, many problems exist with this that still need to be solved. Hopefully they will be figured out eventually. I’ll post more about this later.

I would appreciate any ideas you may have relating to this.

EDIT: See http://dot-p2p.org/ for some people attempting this…

Related Posts:

• A few notes on technology, WikiLeaks, data mining, and privacy
• Datacenters are actually shrines!
• The privacy and security implications of ‘smart meters,’ and the future of energy
• Tunisian government executes man-in-the-middle attack on Facebook, steals passwords and identities
• WikiLeaks deleting and editing cables – you can’t trust them
• Probability analysis of the insurance.aes256 file posted by WikiLeaks