5 ms·
I don't see how that would follow? Lets assume that today that Signal is able to deliver messages securely between users, without maintaining a plaintext assoc
by paranoidrobot 4y ago
I don't see how that would follow?
Lets assume that today that Signal is able to deliver messages securely between users, without maintaining a plaintext association between sender and receiver's phone numbers.
Why would switching identifiers from phone numbers to some other identifier require that they change how that works? (Other than the obvious 'substitute out a phone number for some other identifier' thing)
- tptacek 4y agoPhone numbers are already attached to fairly complete contact databases for users. "Signal Usernames" aren't.
- woojoo666 4y agoCouldn't this plaintext contacts database be encrypted on the server and only decrypted on the client?
- tptacek 4y agoNow you're building Matrix. Look how that worked out: https://nebuchadnezzar-megolm.github.io/ https://nebuchadnezzar-megolm.github.io/
- woojoo666 4y agoSo it's possible, it just takes work, which Matrix is doing right now. Is Signal also working on this?
- tptacek 4y agoThis protocol was deployed for years before anyone even thought to check whether group messaging was secure (we just found out now that it isn't). Is Signal working on secure group messaging? Well, they won the Levchin Prize at RWC for the work they're doing on it.
- woojoo666 4y ago> Is Signal working on secure group messaging? Well, they won the Levchin Prize at RWC for the work they're doing on it. I perhaps should have clarified but I was talking about group messaging without phone numbers attached
- tptacek 4y agoYes, and I'm explaining why that's harder than it looks; for instance, see where Matrix ended up with homeservers.
- woojoo666 4y agoI completely understand the difficulty, but so far I haven't heard of whether or not Signal is even working on getting rid of phone numbers. Also worth mentioning that Matrix isn't just working on encrypting contacts databases. They are also working on decentralization, which is a much more difficult problem. So I'm not surprised that they are running into issues every once and a while (though Signal has had security problems in the past too [1]) [1]: https://thehackerblog.com/i-too-like-to-live-dangerously-accidentally-finding-rce-in-signal-desktop-via-html-injection-in-quoted-replies/ https://thehackerblog.com/i-too-like-to-live-dangerously-acc...
- tptacek 4y agoIt sure seems to be a much more difficult problem, because they're working right now to recover from disastrous protocol breaks that allow servers to decrypt messages in groups. I'm not so interested in what novel stuff they're working on; they don't have the table stakes stuff nailed down. This is secure messaging. It has to work, or it's just LARPing. Can the Matrix team honestly say that their system is ready to handle life-or-death secrets?
- woojoo666 4y agoHackers were able to get at least one person via the Twilio hack in OP's article (which wouldn't have happened if Signal was not reliant on phone numbers). So I wouldn't say Signal is great for life-or-death either. And afaik the Matrix vulnerability we are discussing was not actually shown to have been exploited
- palata 4y agoYes, and they are working on it. It's just harder than it seems, and it is not as important as people seem to think. I value much more the work Signal did on metadata than the username vs phone number debate.