We are happy for people to use PGP, though of course we don't offer an interface to it because that would involve sharing keys with us and defeat the purpose.
We've addressed the legal framework we work in in the linked blogpost, and addressed precisely the scenario you lay out. Governments change, and they love to posture, but the reality is the laws, not the soundbytes.
We would be aware of the outage - have you seen how reliable NYI are?
http://blog.fastmail.com/2014/12/10/security-availability/
They don't have outages.
I'm much more worried about the NSA (or any of http://en.wikipedia.org/wiki/List_of_intelligence_agencies) knowing of a hole in one of the pieces of software that we use and injecting a rootkit with that. Thankfully the same security practices that defend against hackers are also useful here.
See the following for more info - http://blog.fastmail.com/2013/10/07/fastmails-servers-are-in...
Pretty scary isn't it?
All of the above are scary. But this is not without precedence. The UK chancellor couldn't even answer which tax greggs should be paying.
Counterpoint: seems to me that protonmail is doing exactly that.
>We've addressed the legal framework we work in in the linked blogpost, and addressed precisely the scenario you lay out. Governments change, and they love to posture, but the reality is the laws, not the soundbytes.
You haven't addressed any of the points I initially raised. Your blog post skirts the issue quite handily.
>They don't have outages.
Anyone claiming 100% uptime are fraudsters.
We've been there over 10 years, so we actually kind of know, rather than guessing, how reliable they are.
What exactly do you want us to say? We aren't participating in any dragnets. We've said that plenty of times. If you think that the NSA is in our datacentre, tapping our wires. Well, I was there in July and I moved all the devices by hand. I didn't pull them open and audit the circuit boards - but there's a level beyond which reality is impossible to distinguish from paranoia. We don't follow our hardware from the silicon sands through all the steps before it reaches our datacentre. Of course it could have nasties injected in it.
We design our security processes to make it hard for both hackers AND agencies to attack us with a bunch of things. Not running the same switch with VLANs for both internal and external networks is a big thing - switches are a notorious attack vector. Our internal network devices are fully isolated from the external links.
And then - the security agencies reading your email isn't even the biggest risk to your security and your life for 99.9999% of people. We're not going to throw out tons of features that improve their life for a perception of improved security - so end to end encryption isn't a sane response.
They deliver JavaScript to the browser to decrypt messages. While it's true that they don't have access to user messages for as long as they remain honest, that's a profoundly useless property to have. You can just be honest and not try to access messages, which is what Fastmail does. There's no point in handcuffing yourself and holding on to the key.
As soon as a system administrator feels like accessing the messages (either out of curiosity / government order or because an unauthorized user gained administrative control), they can deliver modified JavaScript to the browser that sniffs the password and decrypted content and sends a copy somewhere. This is, very literally, "being able to decrypt data."
The encrypted messages that are sent to non-ProtonMail users require visiting a ProtonMail website to decrypt the message, which has all of the security concerns as above. Furthermore, it's pushing the line of what counts as "email". Fastmail is clearly advertising themselves as an actual email service, where you send content over SMTP and it shows up in the recipient's email client. You can't do that with ProtonMail encrypted messages; the recipient gets a link to the content.
You can build a new protocol that has the properties we'd all want out of a modern messaging system. You might be able to replace email. But it won't _be_ email, and Fastmail is email. If you want Pond, you know where to find it.
And snake oil with a well-designed website, $500,000 of crowdfunding, and a team of PhDs is still snake oil.
The whole thing with email is that it's your own immutable copy. If you can't even read it without going to some website that may or may not exist at a later date, then you don't really have a copy. So you wind up having to extract the plaintext and then keep a copy of that somehow to ensure you never lose access. Sounds like normal email, the hard way.
ProtonMail is security theater. It is simply not possible for a webmail service to maintain a zero-knowledge policy. They can capture your password every time you log in, and use that to decrypt your email if they want to. Or they could send you a backdoored version of their own JS-based encryption library the next time you visit protonmail.ch. (Hushmail got under fire for doing this a few years ago at the behest of the Canadian government.)
At the end of the day, the only thing that is actually protecting your email at ProtonMail is the fact that Switzerland is not yet known to be collaborating with the NSA. If they claim otherwise, they're either lying, incompetent, or seriously lacking in imagination.
If you don't expect the government to follow any rules, then most of these discussions are moot. But what we've been seeing with the Snowden leaks is that the governments are at least attempting to walk the tightrope of legality. Here in the U.S., that means having at least plausible procedures to filter out non-foreigner communications, and going through legal processes like subpoenas instead of simply breaking into computer systems.
These facts have practical concerns. Unless you hypothesize that the U.S. and Australia are putting citizens in prison using secret courts, the government still has to present evidence in a court proceeding, and that has to comport with the 4th amendment in the U.S., and whatever the equivalent is in Australia.
The law in this area is rapidly evolving, but in my opinion as someone with both a legal and technical background, encryption like what Fastmail has gives 4th amendment arguments a lot more teeth. When your data is in an e-mail service that's readily accessible to the service provider, or even data-mined for advertising purposes, it's susceptible to the charge that it's not private information, because after all you're allowing someone else to rummage through it. But if there are protections in place, even if they can be circumvented if needed, that's different. Now you're talking about something that's more like a safe deposit box at a bank or a rented storage unit, which do have 4th amendment protection. They're locked, and only the owner accesses what's inside as a matter of course. The lock can be broken, in an emergency, but that doesn't change the fact that the owner of the facility does not access the contents as a matter of course.
rayiner, how about the reports of "parallel construction" by the NSA? While it's true that the government still has to present evidence in a court, according to recent reports [1] the NSA "tips off" organizations like the DEA.
1 - http://www.washingtonpost.com/blogs/the-switch/wp/2013/08/05...
Guantanamo is still there; the US is putting non-citizens in secret prison without really a court at all.
The other use of the communications metadata is for targeting drone strikes in e.g. Yemen. No due process at all, just a kill list.
Except if Fastmail can decrypt user's data, then they can be compelled to backdoor their system, and also compelled to keep it quiet. Australia is part of the "5 eyes", after all, and from an outsider perspective their government seems particularly hostile/authoritarian.
Failing that legal avenue, would Australia's intelligence services refuse to simply hack their way into Fastmail's servers if the NSA asked them to? Would the NSA refuse to just do it themselves, if it came to that? Surely not.
The president (or whomever) of NYI would certainly not refuse entry to federal agents wanting to install an implant. NYI just has to report an outage, meanwhile Fastmail's servers have grown a new rootkit. And failing that, then it's time to bribe a security guard in the middle of the night.
To me this reads as nothing more than Fastmail trying to provide a false sense of security.