Re:Usenet, authentication, and engineering: early design dec...
by NoGravitas on Tuesday February 27, 2018 @12:50PM
(#16475933)
This may not be exactly what you are looking for, but you might be interested in how signed messages work in OStatus (salmons) and ActivityPub (LD Signatures). The technical details are borderline terrible (especially for ActivityPub), but the upshot is that when a server receives a message from another server that it federates with, it checks the message signature to be sure it's from who it claims to be.
Re:Usenet, authentication, and engineering: early design dec...
by paulie_a on Tuesday February 27, 2018 @07:21PM
(#16479510)
I am apparently stupid when it comes to this but what is blocking that sort of key exchange at scale?
Re:Usenet, authentication, and engineering: early design dec...
by TACIXAT on Tuesday February 27, 2018 @10:02PM
(#16480345)
My thought on this is say there is a network of 1 million peers. You join the network and post in the chat with a signed message. Now there are 1 million clients who want to verify that message and hit you up for your key at the same time. I could see that effectively turning into a DDoS, especially for people on slower connections. Haven't tested it though, so I could be making a wrong assumption.
In a large chat, where you're basically playing a game of telephone, you could sign your messages. At scale it would be infeasible for every user to retrieve a copy of every other user's key to validate messages. Perhaps having optional validation (e.g. you see an important message and you can choose to validate it, that reaches out to the user, gets their key, then checks the signature).
In smaller chats it would be feasible to hold every user's public keys and retrieve them directly. Then allow the messages to be relayed between participants, or encrypted and relayed along the larger network.