5 ms·
Unfortunately, this is the sort of a change that's a red herring for any actual improvements to email security. The use of unencrypted or encrypted link to the
by jcranmer 11y ago
Unfortunately, this is the sort of a change that's a red herring for any actual improvements to email security.
The use of unencrypted or encrypted link to the receiving email provider's MX server doesn't change all that much in terms of who can read the email: it's still sitting in plaintext on the recipient's server (as well as the sender's server), and the group of actors who can sniff traffic on the backbone like that is probably just as easily able to get it from the servers.
The authentication feature is even worse. The problem of spam and phishing isn't that email claims to be from important-service@bigbank.com, it's that email claims to be from "Big Bank" <whoisthis@some.really.shady.ru>. It's been noted before that spammers tend to be the most aggressive at uptaking new "anti-spam" technologies like SPF and DKIM, and this sort of validation feature seems like a prime vehicle for exploitation by spammers.
- skybrian 11y agoYou're arguing that we shouldn't do anything, instead of taking a step in the right direction. Email is an old ecosystem, so it's not possible to make big improvements all at once.
- infogulch 11y agoAnother case where The Perfect is the enemy of The Good. Baby steps, people.
- Zigurd 11y agoIt's a good step, but a very small step. It's been a long time since the Snowden leaks. You're pretty much naked to state surveillance without: a) An open source client that encrypts your mail b) A public key resource with a web-of-trust All the major Web portals/ecosystems could lock the spooks out of email payloads very effectively. That would still leave other issues to solve, but securing payload in flight and at rest seems like the first real milestone that makes a difference.
- jcranmer 11y agoI'm not arguing that we shouldn't do anything. I'm arguing that the presentation of the data is at best meaningless and at worst downright harmful. DKIM authentication is in no way an attestation that it was sent by its sender. Furthermore, from what I can tell, more damage is caused by spoofing that works on the "I use a very similar name which is hard to see the difference" level (e.g., animenewsnetwork.com versus animenewssnetwork.com). Finding and testing solutions to that problem is something that doesn't require changing or deploying anything to new to the ecosystem, and it would arguably bring more benefit than deploying end-to-end encryption.
- jcrites 11y ago> I'm not arguing that we shouldn't do anything. I'm arguing that the presentation of the data is at best meaningless and at worst downright harmful. What do you think should be done to make progress that's better than Google's proposal? Personally I think encouraging all senders to adopt DKIM, transport layer TLS, DMARC, SPF, etc. by displaying auth results from those protocols in the UI is a good first step. It's similar to the push for HTTPS on the web. > DKIM authentication is in no way an attestation that it was sent by its sender. Google didn't mention DKIM in the blog post we're discussing. Are you referring to a related effort? The blog post was about encryption in transit with TLS. > I'm arguing that the presentation of the data is at best meaningless and at worst downright harmful. Google's DKIM solution is as close as one can reasonably get given modern technology. If I send a DKIM-signed email from example.com that passes validation, then it means the following is true: I sent an email through a server managed by the domain owner, that had access to the DKIM key and chose to sign my email. It's not authenticating a person, but it's authenticating that the domain's mail servers sent the message. The attestation is at the domain level, not the sender level. This attestation is still useful though: if the domain owner did not want to allow you to send email from jcranmer@example.com, then it would not accept that email from you or DKIM sign it.
- KirinDave 11y ago> The use of unencrypted or encrypted link to the receiving email provider's MX server doesn't change all that much in terms of who can read the email: it's still sitting in plaintext on the recipient's server (as well as the sender's server), and the group of actors who can sniff traffic on the backbone like that is probably just as easily able to get it from the servers. That's not at all obvious to me. And this sort of backbone sniffing is exactly what we've seen state actors do. What possible downside could rewarding link integrity have on the story for user privacy? A false sense of security? Quite the opposite: that's what we already got in spades.
- magila 11y agoA state actor who can sniff traffic can just as easily serve a secret order to the email server operator to compel them to hand over the data. In fact this is already SOP for governments when dealing with communications which are encrypted in-flight but not end-to-end.
- jamespo 11y agodepends where those servers are based and who owns them
- Kadin 11y agoExcept that's demonstrably not what the NSA was doing; they were sniffing the backbones at the switches, presumably because they didn't want to create any sort of paper trail or get warrants, even secret FISA ones, against particular people. From Google's perspective, it means they have to be involved, however non-consensually, in the process, rather than having it done to them transparently. So I can see why it's a win for them. Whether it really matters to Joe Random User is arguable. What Joe User really ought to demand is end-to-end encryption rather than transit encryption, but he's unlikely to do that because key management across multiple devices (and who wants to read their email only on a single device?) is a pain and the companies that would be able to make it less of a pain aren't exactly incentivized to do so.
- thrownaway2424 11y agoThere's a lot of supposition in your post and no evidence in favor. First of all what makes you think Gmail stores user data in plain text at rest? Secondly think about your definition of "backbone" networks. Google's own network is extremely large in scope, and generally user traffic lands in Google's network at a point very close to the user. For example if a user in Japan sends mail to Google, that connection will be terminated in Japan, and backhauled across Google's network to wherever the actual servers are located. Google intentionally shrinks the scope of network surveillance by minimizing the network distance between the user and Google's frontends.