Hacker News Reader

Comment by nextos | Hacker News Reader

I wonder what happens with Signal as it's US-based.

Signal may or may not be ok at the moment, but not in a couple of years (it has all the power to silently push an update with a backdoor). Among all the popular messaging apps only Telegram is in a position to not cooperate with five eyes.

Sure, but the Signal source code is open source. I've been compiling it myself for years on Android (and signing the binary with my own key). They can't silently push a binary from the Play Store and overwrite my binary.

They still can for everyone else, probably everyone you talk to over Signal, so not much of a protection. Pretty much the only way to avoid that is for Signal to give up that control, don't supply Signal directly and let others build Signal from sources and publish in independent from Signal package repositories, i.e. f-droid, OS package managers on Linux, etc. Even if they don't federate, it will at least make their end-to-end encryption sound, not useless like it is now when everyone uses Signal supplied binaries.

Presumably, for most of us here, even if we compiled it ourselves, most of the people we'll communicate with won't.

Dude, I can only imagine what you are doing that you need to be so secretive lol

I think "trust, but verify" and "verifiable security" are good axioms.

I use to defend Telegram but they are technically UK based AFAIK.

They have a legal entity registered in the UK which they can drop any time.

I thought it was Berlin.

Whenever we talk about web encryption, people make the (valid) point that needing to trust the server to deliver the encryption mechanism greatly reduces the benefits of clientside encryption.

I agree with that analysis, but it's not clear to me why we don't have similar levels of skepticism about auto-updating desktop apps. Signal in particular uses a third-party software repository, so if it wanted to push a malicious update, it wouldn't even need to sneak it past package maintainers.

Package signing protects you against developers with bad personal security practices, because it makes it harder for a third-party to MITM their apps. But it doesn't do anything I can see to protect you from a developer that turns malicious in the future.

Unfortunately, unless things have changed since last time I tried telegram, it doesn't encrypt chats by default, which is a huge problem for any movement. OpSec is really hard, even for those who know what to do. (Operators of Silk Road 1&2 were both ha taken down by a failure in OpSec somewhere.) Security researchers have cast aspersions on MTProto the protocol used. The truly paranoid I know, for whom Signal and Telgram don't go far enough, use Wickr, which chooses to more make sacrifices in usability in favor of security, but they are few and far between.

I'm not sure if I completely agree with the silently part. Doesn't Signal have reproducible builds? In theory, any binary they release can be checked to see if it corresponds to source code.

Whether anyone is actually doing that is another question. And there is no technical reason app stores couldn't send a special backdoor-ed build to select list of users under surveillance if government forced them to. (They can target sets of users for staged roll-outs, beta programs, etc.) Which defeats the notion that one person watching can detect it for everybody.

On the other hand, it's possible to do stuff with smart phones at the platform level. Whether through vulnerabilities or some platform capability (updates, etc.), it may be possible to have a backdoor-ed binary that looks to the user like it is the regular binary. That's not a capability that Signal has, but it is a capability that might very well exist.

Regardless of all that, for users who have auto-updates enabled (most users), even if Signal can't silently push a backdoor-ed update, they can unilaterally push one. You could wake up tomorrow with a different version that has a backdoor, so even if you can identify backdoor-ed binaries, you have to turn off updates or verify the binary every time you open the app.

Assuming for a moment that we can trust our smart phones at the platform level not to lie to us, the problem of targeted backdoored builds could be mitigated somewhat if the platform implemented Binary Transparency in a clever way:

https://wiki.mozilla.org/Security/Binary_Transparency

When you install an app (whether through an app store, or side-loading) the app should state the location online of an append-only log that lists all the releases (with timestamps) for that app. The phone OS could periodically check to see if an upgrade is available, and security researchers could check that the log doesn't contain references to versions which aren't available to the public.

Ideally there should perhaps also be a way for users to anonymously report which version of any app they are using, so that people with particular security concerns could configure their OS to only update an app after, for example, 50% of users have already installed the update.

> Among all the popular messaging apps only Telegram is in a position to not cooperate with five eyes.

Telegram is subject to US coercion as much as any US company: both major app stores are US-based, and without app store distribution, a product might as well be dead as far as the masses are concerned.

Maybe @moxie could enlighten us.

Use signal, use tor.

> Use signal

I would, except it requires a mobile phone number.

Nothing. Whoever is interested, has to get hold of the phones to access the messages.

Historical messages, yes. However an app maker can be coerced into adding a hidden user that can participate in chats and receive all future messages decrypted. It doesn't require any special crypto hacks for that to work.

Signal code is open. You can build it yourself and compare that build with whatever you download from the app store. So it's not that simple.

I don’t think the iOS App Store allows for that sort of checksum comparison.

Reply on news.ycombinator.com