This actually happens a lot. Most people most of the time don't painstakingly type in h-t-t-p-s-:-/-/... but rather just the domain name and the server on the other end returns a (plaintext) redirect to https. Which is why sslstrip works so well.
Sure, we have bookmarks, HSTS, tell people to look for a lock icon in the address bar but it's far from being a solved problem.
If you're using other browsers, the problem still exists but a large majority is covered by this.
Banks have the ability to do it, so I wouldn't exactly pin it on a new solution being needed. If they won't adopt pinning or HSTS, then why would they adopt anything else?
For the record I do not think it is a final solution (what is). I do often have mixed feelings about 'the perfect being the enemy of the good'. With STARTTLS my feelings aren't as mixed. A measurable improvement to passive surveillance for minimal changes and no new infrastructure. Swell.
Again, not going to condone it as a panacea - but it's never advertised itself as one.
Let's keep using it until there's something better. And let's get furious at ISPs that strip it (or modify our traffic in any significant way).
STARTTLS, a protocol used to negotiate SSL/TLS in some plain text protocols, is problematic if it isn't enforced. Some software stupidly abbreviates STARTTLS to TLS in the GUI, which is a source of constant confusion.
STARTTLS is simply a method for establishing a TLS connection through the same TCP port as the backward compatible connections that don't use TLS. It's the job of TLS to provide authentication, and thus to protect from MiTM--it would be braindead to have a second authentication layer beneath it to "protect STARTTLS against MitM". If you establish a "direct" TLS connection, the TCP beneath also doesn't protect against MitM, obviously.
That some clients will accept both TLS and non-TLS connections is completely orthogonal to STARTTLS, and has its roots in backward compatibility. Good clients should provide a way for the user to enforce TLS, though.
And yes, as others have pointed out, HTTP(S) does actually work exactly that way: Unless the user asks the browser to enforce TLS (by typing https://), the browser will connect without TLS, and if the server prefers TLS, it will redirect, subject to the same MitM as opportunistic STARTTLS.
Also, I think it's kinda confused to think of your customer's traffic as your own. How does that work? Are telecom companies also allowed to listen to "their own" calls that customers make, and the post office to read "their own" letters that customers have written? Also, you have considered that your ISP's ISP then presumably also could consider all your traffic their own? So, there essentially wouldn't be anyone left who couldn't claim the traffic to be their own, so MitMs wouldn't be possible simply by definition?
https://github.com/EFForg/starttls-everywhere
I'm not sure one could say that this makes STARTTLS non-broken, but it will help mitigate the MITM risk.
Exim will soon also be getting DANE support. Not sure if it is being worked on by other mail server teams or providers.
In brief,
- DNSSEC enables and facilitates reflected DoS attacks by amplifying attackers' bandwidth;
- Using DNSSEC allows anyone to enumerate any of the zones of your subdomain, effectively turning on public DNS transfers for anyone who asks (see the paper for attacks against NSEC and NSEC3);
- Most importantly, no common resolver validates or enforces the validity of DNSSEC records. Chromium closed the pull request as WontFix: https://code.google.com/p/chromium/issues/detail?id=50874 and Mozilla have no current plans to implement it: https://bugzilla.mozilla.org/show_bug.cgi?id=672600
That's a problem today, without DNSSEC, and requires fixing regardless, whether or not DNSSEC becomes wide-spread.
- Using DNSSEC allows anyone to enumerate any of the zones of your subdomain, effectively turning on public DNS transfers for anyone who asks (see the paper for attacks against NSEC and NSEC3);
I thought NSEC3 fixed that, but I wouldn't be surprised if there were attacks against it. I don't think that's an insurmountable problem.
- Most importantly, no common resolver validates or enforces the validity of DNSSEC records. Chromium closed the pull request as WontFix: https://code.google.com/p/chromium/issues/detail?id=50874 and Mozilla have no current plans to implement it: https://bugzilla.mozilla.org/show_bug.cgi?id=672600
Chicken and egg. shrug There's no reason an OS can't come with such a resolver built in. Personally I take the 60 seconds or so required to do an "apt-get install unbound" to get this functionality whenever I build a system.
What's the definition of a MitM? If you're trying to talk to Google, and your ISP is modifying what you say, I would consider that a MitM. Wikipedia mentions a router MitMing its users. I don't see how that can be considered a MitM but not this.
STARTTLS is sent via the unprotected SMTP channel. The biggest problem with this isn't the lack of encryption, but more the lack of MitM protections.
Imagine if when you wanted to connect to your bank over HTTPS you first had to send a HTTP request which asks permission to "upgrade" to HTTPS. Obviously any bad guy worth their salt is going to intercept that unprotected HTTP traffic, strip out the STARTTLS, and then leave you using the HTTP channel they can monitor or alter.
So while, yes, in this case ISPs are likely "MitM"-ing their own traffic (which by definition isn't a MitM), it still leaves you open to bad guys on your side of the connection (e.g. the insecure WiFi problem). But even if ISPs weren't someone else could be.