Post 9wLSdXPBlDXfomgzvE by christianbundy@social.coop
(DIR) More posts by christianbundy@social.coop
(DIR) Post #9wFIpsMMfXpKVg5gsi by solderpunk@tilde.zone
2020-06-19T16:06:03Z
0 likes, 0 repeats
@xj9 Has the idea of using ssb for distributing TLS certificate hashes ever been discussed in that community?
(DIR) Post #9wFJmaJjOUXpdaqz8S by 361.xj9@social.sunshinegardens.org
2020-06-19T16:16:41.245713Z
0 likes, 0 repeats
@solderpunk I think @christianbundy might have touched on this idea somewhere? he's been experimenting with ssb over http recently.I wonder if something like shs could be used as a secure capability based alternative to finger. OG shs is a muxrpc protocol though. https://dominictarr.github.io/secret-handshake-paper/shs.pdf
(DIR) Post #9wFOrlbnL8ujsMaplI by solderpunk@tilde.zone
2020-06-19T17:13:36Z
1 likes, 0 repeats
@xj9 @christianbundy Thanks for the link to the paper! I'll read it after the TLS TOFU paper I printed earlier today!
(DIR) Post #9wFmpSuwWSxRaFrEyu by christianbundy@social.coop
2020-06-19T19:44:55Z
0 likes, 0 repeats
@xj9 @solderpunk it me! happy to answer any questions.summary: the current replication uses trusted authentication, which is nice if you need to verify both parties, but since our messages are already encrypted I don't think it's necessary. we can just say "here's my TLS cert, hit me on example.com" and I think that's Good Enough.
(DIR) Post #9wFmpUhtrviH8Nlqvg by solderpunk@tilde.zone
2020-06-19T21:42:04Z
0 likes, 0 repeats
@christianbundy @xj9 Oh hai! Sorry, I've come into this with a pretty minimal background in ssb and don't immediately understand what it is you've been experimenting with. Is there any kind of writeup anywhere I take a look at to try to get up to speed?
(DIR) Post #9wLSdXPBlDXfomgzvE by christianbundy@social.coop
2020-06-22T15:24:07Z
0 likes, 0 repeats
@solderpunk @xj9 Hmm, no writeup so far. Maybe I can summarize: SSB messages are signed and can be validated with the author's public key (https://ssbc.github.io/scuttlebutt-protocol-guide/#message-format) but conventionally those messages are replicated with a homebrew network protocol.I'm lazy and boring, and so I've been trying to switch to HTTP replication instead. My HTTP server is < 100 LoC (https://github.com/christianbundy/http-ssb) and there's a demo too (https://daily-alluring-robe.glitch.me/).There's no cert pinning right now, but it's possible to add!
(DIR) Post #9wTSFMCJx0bjiBAaFk by solderpunk@tilde.zone
2020-06-26T11:57:28Z
0 likes, 0 repeats
@christianbundy @xj9 Thanks a lot for the clarification! I understand now what you're trying to do. I think it's quite different to what I was thinking about: using SSB's "append only log" concept as a way for people to securely broadcast the fact that a string of self-signed certificates were, in fact, all generated by the same entity as part of a legitimate key rotation schedule.
(DIR) Post #9wTXQbrQYWp1FIa8zw by 361.xj9@social.sunshinegardens.org
2020-06-26T12:55:28.946881Z
0 likes, 0 repeats
@solderpunk @christianbundy we could try doing that with rlog, but the format needs some revisions before I'd want to write a spec. i've been working on a new design on paper that depends on TLS instead of libsodium and uses a new TLV encoding variant based on the original TSV format.
(DIR) Post #9wTn0s5JDCD9Ihp8wy by christianbundy@social.coop
2020-06-26T15:46:26Z
0 likes, 0 repeats
@xj9 @solderpunk oh, interesting! so you'd just publish certs and have users subscribe to cert changes?
(DIR) Post #9wTn0sNO7yfmCmHa88 by solderpunk@tilde.zone
2020-06-26T15:50:06Z
0 likes, 1 repeats
@christianbundy @xj9 Exactly. Kind of using SSB (in conjunction with TOFU for initial connections) as a decentralised alternative to the CA system. The, uhh, feed ID, or whatever the right term is, for future cert announcements could be put in one of the fields of the cert Subject (hmm, or even Issuer) Common Names.