Hacker News Reader

Comment by xyzzy123 | Hacker News Reader

This is cool, but it doesn't actually help you with mass surveillance unless the other side of your conversation is doing that too. What point running your own email server if your friend is on hotmail?

Also, even if you both run your own MTAs STARTTLS does not help much except to reduce the scope of the passive-only surveillance dragnet, because the configuration does not require signed certificates.

Finally, if both parties are technical enough to run their own email servers, um, why not just use GPG?

From a technical point of view, I put this squarely in "fun, but a boondoggle" territory.

View on HN

> Also, even if you both run your own MTAs STARTTLS does not help much except to reduce the scope of the passive-only surveillance dragnet, because the configuration does not require signed certificates.

Exim and Postfix (and hopefully others to come) are both working on adding DNSSEC and DANE support right now. Which means in the not too distant future it will be possible to publish cryptographically signed data in the DNS that your MX has TLS support, and what the fingerprints are.

> Finally, if both parties are technical enough to run their own email servers, um, why not just use GPG?

OpenPGP encrypted email leaks all sorts of information/meta data.

The metadata problem is so hard that if you actually solve it you are not using email anymore. This is why I'm saying it's a boondoggle. Even in the best case where my mother and I are running our own STARTTLS secured MTAs, oh wait, timing and message size.

GPG gives you actual privacy but minimal resistance against traffic analysis. All this other stuff gives you NO end to end privacy, AND minimal resistance against traffic analysis.

Messaging is a weird special case and GPG is behind the times (because people are coming to realise that repudiability and FS are important), but it's still a heck of a lot better than even working TLS slapped on SMTP (which doesn't even currently exist).

The security of email will never be perfect; that is obvious. But there is still legitimate and real value gained from making small incremental improvements to it. If you can hide the To/From/Subject of a message from a passive or even active MITM (because of DANE), then that's a good thing.

In this situation, you are the one who gets the national security letter or warrant. Also, it requires a warrant because it's on your personal property. They don't need a warrant to search your ISP.

At least in theory this shields who you're talking to, somewhat. But I think a radical rethink like Pond is perhaps a better way forward:

https://pond.imperialviolet.org/

I think we're talking about different orders of magnitude of weirdo here :)

At this point the only people I know who could or would use Pond are people who can compose crypto primitives on their own anyway.

Also if we are going to go all Pond and do messaging from the ground up, we need to acknowledge that the global passive adversary is real.

EDIT: after replying to rickisen below, I'm starting to see your point. Maybe we need Pond or bitmessage or a resurrected mixmaster if real people think TLS solves traffic analysis.

metadata

Let's think about this more closely.

Where are you hosting your email. Are you keeping it on say an AWS instance? OK flag that, all the plaintext would be there, and also, duh, it's still obvious that it's your communication because you're connecting to it, and maybe even paying for it.

OK, you host it at home on your DSL line? OK, so plaintext is harder, but oh wait. Yeah, your IP.

TOR hidden service or something? Good luck getting SMTP to work over tor.

So tell me again how TLS is going to help with metadata.

Reply on news.ycombinator.com