Hacker News Reader

Comment by woodruffw | Hacker News Reader

Why is “owned by the users” (whatever that means) the quality you’re testing cryptographic protocols against?

Signal's has buildable source code on the client side and all. However, the server code is closed source and the developers have made disquieting changes like uploading most users' contact lists to the server protected only by Intel SGX that is known to have vulnerabilities exploitable by a nation-scale actor.

The point of E2EE is that you don't need to trust the server that you're talking through. A secure E2EE scheme is secure in the presence of an actively malicious server.

I think SGX is a red herring in this context. You can find it risky or not, but it's (1) not material to Signal's E2EE property, and (2) materially better than email, which sends all contact metadata in plaintext, regardless of whether you try and encrypt your message bodies. Bringing it up in this context seems like a category error.

Is this accounting for the fact that they kill people based on metadata? More technically complex and unreliable protocols such as SimpleX attempt to also protect the metadata.

Is what accounting for that? If you're worried about metadata leaks, you shouldn't be using email, full stop.

By "metadata" GP is presumably talking, as I was, about Signal uploading most users' contact lists to the server with weak protection.

The point is that if the comparison point is PGP-encrypted email, which has no protection for any metadata and passes it on the wire in the clear, raising concerns about Signal’s metadata protections makes no sense.

In the context of this whole big thread, the point is that neither infamously leaky PGP nor Signal keep metadata secure from nation-level actors.

Reply on news.ycombinator.com