From their blog post some days ago, I thought it did just that:
> Unlike a standard HTTP proxy, connections to the Signal TLS Proxy look just like regular encrypted web traffic. There’s no CONNECT method in a plaintext request to reveal to censors that a proxy is being used. Valid TLS certificates are provisioned for every proxy server, making it more difficult for censors to fingerprint the traffic than it would be if static self-signed certificates were used instead. In short, everything is designed to blend into the background as much as possible.
They should probably make that post less reassuring and list the exact risks.
Given its broad user base, it wouldn't hurt for Signal to clearly state "we can't keep the fact of the communication private, only the contents".
That would short-circuit attempts to gain notoriety by pointing out obvious facts and calling them vulnerabilities. It's also common sense for anyone who knows their way around a puter, but that's not Signal's median user.
"Privacy" products are a market for lemons, and Signal's public messaging should strive to insulate its users from FUD.
Everyone in my country of residence uses Telegram; not because it's secure, but because for non-serious chats, it's convenient.
This is a statement/truth that I think a LOT of people don't quite get; Pavel Durov and his team might push that Telegram is secure, but no one uses Telgram for security because nothing about its security ensures a circle of trust.
This is true for any messaging app. The general consensus I've encountered is that any application you can readily pull from the AppStore/GooglePlay, so can adversarial persons. If they really want to target you for some reason, it's as simple as getting your friend and unlocking their phone, knowing the the same protections that people in the United States have don't apply world wide.
Signal et.al., can have the world's most amazing crypto, but it means nothing if the person behind the unlocked phone is an adversary, and I think a threat-vector that is missed in the 200+ post discussion here is that in many parts of the world, __this is a real threat vector__. A password/second screen/whatever is not a guarantee of protection! It's a speed bump, and how resilient you are to the person driving over the speed bump is how effective that bump is.
So, for the article when I read persons concerned about the statements made by Signal and not outlining the threat factor for countries where the Circle of Trust might literally be a matter of life or death, yeah, I side with the concerned persons. Signal has impressive tech, but again, that tech means __nothing__ if the person behind the unlocked device is adversarial.
So I feel like I am missing something here.
I don't want to minimise empathy towards living under such a society but if The Adversary can obtain your friends unlocked phone, what do you want to have happen?
A sham-password for Signal that sends out a "I've been caught." message to all contacts? So you can "fat finger" the password in front of your captor?
I don't understand. How would you circumvent censorship of the protocol without obfuscating the protocol?
It seems to me that signal has never claimed to be able to hide that it was being used... until now?
But thanks for posting the thing from Moxie, it does sound quite reasonable.
What would be useful to me and presumably other HN readers is a clear summary of the tech involved, readable by an audience who is technical but not security/circumention experts. Like, the people complaining could be spending time on that, to educate users and developers, instead of doing... whatever they are doing. That seems to have turned into a much less interesting argument about etiquette or something.
So I guess they're trying to pop up as many endpoints as possible to circumvent that.
I don't know the details about the network block though, so I might be mistaken. But the nginx config in that repo is purely a TLS proxy. Nothing magical happening there at all, just an entrance node to the main signal network
End-to-end encryption means content privacy but not necessarily meta-data privacy or anonymity.
That is a very odd statement.
I do wonder how Moxie knew this guy knew that GH is not used for discussion. Maybe the only way to tell is to see that there are no other active discussions?
Why isn’t Signal just a Free and open source, infrastructure-less p2p solution? Maybe the goal isn’t really security or privacy after all...
Great question! It's a good way to make it easy for general-purpose users with limited technical expertise to adopt, use, and find one another.
> Seems pretty weird for a “secure” program.
You're right! It's definitely weird, but it's also understandable as a tradeoff in favor of less technically adept users. It's not one I'm in love with, but I think it makes sense.
> And why does it use AWS? Isn’t that subject to all kinds of privacy risks including National Security Letters?
The risk from NSLs depends a lot on what is hosted. If it's opaquely encrypted blobs, there's minimal risk. And where could things be hosted that wouldn't be subject to privacy risks from a government of some sort?
> Why isn’t Signal just a Free and open source, infrastructure-less p2p solution?
That's such a good idea that Signal is already a Free and open source solution!
That said, nothing is ever actually infrastructure-less, just like no data store is actually schema-less. There's just explicit infrastructure and implicit infrastructure. Implicit p2p infrastructure is not immune to governments or NSLs, and is often subject to more by virtue of being in more countries.
Personal insults aren't called for. Please stop.
And even though I disagree with the focus on phone numbers and wish they'd prioritized a model that makes them optional, I do understand the network effect argument for including it. Kind of annoying that the alternatives that do it better have a hard time, but I also have to admit that it proves Moxies point to a degree.
Different people have different threat models and I think asking for a phone number for the reasons posted above is acceptable for most people.
Wifi is even worse, not better..
You can find any number of infrastructure-less p2p solutions. The number of users they have compared to Signal might be illuminating.
If you think that, just by making authorities know your phone number is registered on Signal is dangerous enough for you to be arrested, you should not use Signal.
Signal, like any other software, can not solve political, or dictatorship. Signal is a chat app, not a magical tool, even if it is helpful for those objectives. That's what we mean when we says "security is layered".
So, if your government have unlimited resources (that is to say, they can simply arrest and sentence you if they *think* you *may* using Signal, Telegram, Whatsapp, Tox chat, ..., without judicial review), then maybe Signal is not your biggest problem.
Whether or not AWS is risky, I don't think signal has any increased risk hosting their infrastructure on it vs. any other service. The whole point is that comms are end-to-end encrypted from handset to handset, and so any data in Amazon's hands is encrypted.
I think that says it all.
I'm also a bit concerned that "security researchers" don't seem to understand the threat model. Signal has never claimed to be able to hide that it was being used. The TLS proxy is only meant to help circumvent censorship, not obfuscate its protocol. And indeed, as a temporary solution, it's not ideal even to circumvent censorship. But they're apparently working on something better, and all this distraction is not helping.