3 ms·
"People have tried to fix email but nothing as ambitious as TLS/HTTPS has been. And getting people to use a more secure platform built on top of HTTPS is likely
by jerf 10y ago
"People have tried to fix email but nothing as ambitious as TLS/HTTPS has been. And getting people to use a more secure platform built on top of HTTPS is likely a non-starter..."
Makes me wonder if it's time to metaphorically pack it in, and respecify the SMTP infrastructure on top of HTTP(S), precisely because that seems like the only way we're going to get cert security with email systems. As long as it's an optional add-on to SMTP it seems it just isn't going to be added on. (Of course SMTP wouldn't go anywhere right away; I'm talking about a real process with transition times and such, not a mystical one where this would one day replace SMTP in a big bang.)
I mean, there's a loooot of i's to dot and t's to cross betwixt this little comment and an actual standard, but conceptually it doesn't seem too difficult. SMTP is conversational standard but it seems like we've probably got enough negotiation tech in HTTP to pull it off in a request/response manner nowadays.
- Xylakant 10y ago> Makes me wonder if it's time to metaphorically pack it in, and respecify the SMTP infrastructure on top of HTTP(S), precisely because that seems like the only way we're going to get cert security with email systems. Transport security is not the actual issue. SMTP over TLS has been around for a while and is fairly well functional. The problem is attaching an identity to the senders email. That's what S/MIME and GPG/PGP do, but the actual real-world problem here is that you need to somehow certify that the sender is the right person. So you can either have a centralized set of authorities (S/MIME) or Web of Trust (GPG/PGP). Neither option actually scales. Some countries started issuing certificates in their ID cards, but given that other countries don't even have ID cards, this is obviously not going to fix this either. HTTPS has the same problems in principle, but it only needs to certify a comparatively small number of entities (web servers) as opposed to actual users.
- jerf 10y ago"Transport security is not the actual issue." From what I hear from security folks, transport security is still an issue. You can negotiate up to TLS easily in SMTP, as long as you don't care about certificate validity. But without caring about certificate validity, MITM is still quite possible.
- Xylakant 10y agosure, there are still providers that don't offer TLS, but my point is that fixing TLS doesn't even begin to tackle this actual problem. It's an orthogonal problem. This issue is about authorization/authentication, not about transport security/MITM attacks. PGP and SMIME, even when used for signing only will protect against this attack while even fully deployed TLS will not.
- semi-extrinsic 10y ago> Transport security is not the actual issue. Depends on what you mean by transport security. I agree TLS solves important aspects, but I can still pretend like I am xylakant@example.com without issues, since most domains (especially corporate ones!) don't enforce strict enough SPF/DKIM/DMARC policies. If anyone can pretend to be any user, I'd say transport security is lacking. In Thunderbird I have a DKIM validation plugin that shows the status below the subject line. I've seen e.g. large global oil companies that don't have any DKIM on their emails in 2016. That means anyone could send emails as ceo@oilcorp.com (or better yet (cfo|ITsupport|HR)@oilcorp.com) to employees of oilcorp and it would look genuine.
- Edmond 10y agowe are working on addressing the true-ID problem with a privacy preserving trust relay protocol: https://github.com/CipherTrustee/Trust-Relay-Protocol-API/blob/master/README.md https://github.com/CipherTrustee/Trust-Relay-Protocol-API/bl... Basically a scalable way to verify identities in a federated manner.
- Xylakant 10y agoI wish you best of luck, but I'm afraid this is a pipe dream. You can't tie a real person/institution to an identity in a privacy preserving manner which is what's required here. This is fundamentally a social problem and I'm afraid no technical solution exists. Technical tools may help, though.
- sverhagen 10y agoI'm all for technical solutions to this problem, but after reading the original article, I can't help but pointing out that this scam was social engineering, not technical engineering.
- jerf 10y agoA technical solution to transport security doesn't solve very many problems on its own, either human or social; it's just a prerequisite to truly solving most of them.
- dsr_ 10y agoYou might be interested in Dan Bernstein's IM2000 proposal, which centers around the idea that instead of sending email to a local server which forwards it to some number of other servers before it lands in an inbox, the local server hangs on to the body of the email and just sends a notification that it's available to be read. This helps reduce the spam problem, because a mail sender needs to be contactable in order to read the content. Serving up the mail with https is a no-brainer. You still have reputation problems and certificate authority issues, but the value of a botnet to send spam is greatly reduced.
- nkrisc 10y agoThat's pretty interesting. Once the recipient contacts the sender to view the contents, presumably their client could also download the contents to have it locally available, no? Would the retrieval of the email have to be user-initiated?
- dsr_ 10y agoThe retrieval has to be user-initiated, but there's no reason that the user can't automate it, especially for senders on a whitelist.
- sverige 10y agodjb has made some pretty useful stuff over the years. I hadn't heard of this proposal, but it makes complete sense.
- nkrisc 10y agoThat's cool. So for known senders the end user's experience would be no different than it is today.
- astrodust 10y agoIt just moves the failure points around. One day you can't get email content from anyone at Gmail, but everything else is okay.