They share it with you. Their public key is generated by them. You encrypt a message to them with their public key. They use their private key to decrypt it.
I want to add before I get completely roasted here, that this is intentionally reductive. Signal actually uses a much more interesting multikey sharing algorithm, double ratchet. This uses onetime keypairs, and really is worth reading about.
I’m not following. In the WhatsApp case, yes, because we can’t see how those keys are managed. In the Signal case, we can. So the centralized server has zero impact on the privacy of the message. If we trust the keys are possessed only by the generating device, then how does the encrypted message become compromised?
I’m not talking about anonymity, only message privacy. No different than any of the other proxies or routers along the way. If they don’t have the key, the message is not readable.
Sorry but you’ll need to hold the L on this one. If I encrypt a message with public key material for which the only private key material that can decrypt the message is in only my possession, it doesn’t matter if the message passes centralized servers.
I’m not trying to be rude, that’s just how it works.
Actually great questions. Yes and no. There are vulnerabilities if the private key leaks, but public keys are just that; perfectly okay public in any hands. You only encrypt data with it.
What makes the Signal protocol so awesome, and other algorithms like it, is that it reduces the threat surface area further by using onetime keys. So even if your key is leaked, it cannot be used to decrypt old or forthcoming messages as the keys have already ratcheted to the next pair.