https://github.com/signalapp/Signal-Android/issues/127#issue... (2013)

src: https://news.ycombinator.com/item?id=11672199 (2016)

alternative: https://signal.org/android/apk (2017)

You can get a phone number from https://jmp.chat/ if you like. The signup process can be done entirely over Tor.

If you don't use it beyond the trial, it's like the public payphone option mentioned by another commenter - someone could take your number. But if you choose to get a paid account (which, among other methods, can be acquired using Bitcoin, Bitcoin Cash, or a prepaid gift card purchased with cash), then the number will be yours. JMP is probably the most anonymous way of getting a phone number.

I'm not sure of your assumption that lack of total anonymity implies no privacy. They are independently important concepts. You can have privacy (no knowledge of information shared) without anonymity.

That's true, but I believe the concern would be that there's information just in knowing:

1) what's your number, and 2) with whom you connect or communicate.

That is, there's still the danger of social graph analysis: "Oh look, this person's communicating with a known journalist!"

The phone network can do that just by correlating traffic to devices, it doesn't need any help from the software running on those devices.

If the target use case was people that could reasonably do the job of limiting the information just the use of their devices leaked, it might matter. But the target use case is replacing SMS and the like, so it really doesn't matter (except that people want to pretend that the use case is something other than replacing SMS).

https://medium.com/@thegrugq/signal-intelligence-free-for-al...

Of course there is. That's why one of the cleverest parts of signal is the engineering to invalidate number 2. Moxie and signal are the world leaders in trying to make it impossible for the service to know who you communicate with... Even to the point of using the Intel secure enclave to audit the server software and validate that it doesn't peek.

That's not the point. Why doesn't Signal provide the option of using a pseudonym? The model is broken from the perspective of people who care about privacy.

F-Droid wants to compile their own binaries, and Signal/OWS/Moxie want to provide checksums and the like for Signal binaries.

That's an argument for reproducible builds; building the same source should give the same checksum.

The funny part is that Signal and F-Droid both have their own reproducible build system, but they’re incompatible (part of that is that Signal requires proprietary code in its binary)

Can you elaborate more on this? What is the proprietary code used by Signal?

They link several proprietary libraries into their APK, including GCM (which has its own internal analytics package)

Theoretically, but when you throw in things like build systems often not being deterministic, minor versions of dependencies changing, different OS or slightly different OS version with different libraries; there's a multitude of places to throw the final binary off by a few bytes or more and end up with a different checksum.

Signal wants to distribute a binary with a checksum. Once the checksum is different all bets are off, that's why it's not in F-Droid

As if reproducible builds hadn't been done before. If Debian can get to building 80% of their packages reproducibly[1], the communities around Android can get there too. Luckily, it's being worked on.[2]

Now the question is: (when) will this be supported by F-Droid?

[1] Scroll down for a big graph https://wiki.debian.org/ReproducibleBuilds

[2] https://github.com/signalapp/Signal-Android/wiki/Reproducibl...

F-Droid has supported reproducible builds for years: https://f-droid.org/en/docs/Reproducible_Builds/

The real question is, when will Signal finally support it?

Nice, thanks for pointing that out!

So why don't they just also list checksum of the F-Droid binary?

They don't trust F-Droid.

Yeah, figured. They seem very inconsistent in applying their trust. At times they'll do strange things like build app on Chrome Apps platform / mandatory phone ID and on other times they'll make user-hostile decisions like hijacking SMS messages and refusing to publish to F-Droid due to "security".

The end result is an app that keeps shooting itself in the foot and being beaten by Messenger and WhatsApp.

Some more detail: https://github.com/signalapp/Signal-Android/issues/282

The reasons why have been given by others, but you might also like to know that you can download the APK yourself and that it includes its own updater: https://signal.org/android/apk/

The phone number is the contact discovery mechanism on Signal. If you don’t want that, you don’t want Signal.

Me and the persons I want to communicate with are perfectly capable of finding eachother through other means - ids, qr codes, ...

At the same time, maybe I don't want everybody who has my phone number to see that I am on Signal.

You can signup using the number of a public payphone if it makes you feel safer, with landline verification. (Of course, you expose yourself to having the account hijacked by anyone who figures it out and has acces to the payphone).

But it won't improve your anonymity significantly, unless you also use it over TOR.

> I have to sign up using a phone number?

Use a voip phone number if you want, you only need it to get started.

Search for Noise. I believe Copperhead OS is working on a fork that's compatible, but doesn't rely on Google Play Services.

The swiss messenger Threema can be used fully pseudonymously, but is paid (Bitcoin available) and not open source: https://threema.ch/en

Telegram is available on F-Droid. It's similar to Signal with more functionality and greater ease of use - https://f-droid.org/packages/org.telegram.messenger/

You need a phone number that can receive texts for the initial setup, but once you're set up people can add you by @username and never need your number. Stuff like https://www.textnow.com/downloads works just fine for the initial text. Once you have a single device set up, it messages your existing devices rather than sending SMS when you try to connect another device.

One of the main people behind Signal actually tried to spread a bunch of FUD about Telegram years ago, saying the crypto was weak, but it's really not. No working POC code was provided to decrypt anything, just FUD.

Protocol details here: https://core.telegram.org/mtproto They just released MTProto2 in the last year.

Telegram isn't remotely similar to Signal. Telegram communications aren't encrypted by default, and Telegram group chat messages aren't encrypted at all.

This is 100% false. EVERYTHING that goes over the wire is encrypted, always, just like when you're on a TLS website such as your bank.

Group chats aren't end-to-end encrypted, and 1 on 1 chats are only end-to-end encrypted if you make it a Secret Chat.

Did you really think they were talking about SSL in this context? Of course they meant E2E.

To say there's no encryption AT ALL when it's fully encrypted over the wire is still false. Not having E2E encryption is different than not having encryption AT ALL.

They are encrypted in the same sense that the Sesame Street website is encrypted.

So, it's similar to Facebook Messenger rather than to Signal?

(Actually I think Messenger might support E2EE group chats, but I'm not sure.)

All crypto is weak until proven otherwise. Telegram never received a good review from cryptographers. The fact that no POC was provided may just as well mean no cryptographer cares enough to find a bug.

No, he said the crypto was weird (which it is. Who the eff uses IGE mode?) and that their competition to find vulnerabilities was bullshit and would be secure even using crypto primitives that are known to be weak.