End-to-end encryption is an extremely nice feature as it lets us make the assumption that messages cannot be read by a middle man even if 100% of the employees are corrupt and the messages pass right through FSB and NSAs networks twice.
But, as you point out: for most people this falls flat in most cases once you are up against the big guys and they have decided to get you.
That does not mean that it is useless though: most people won't get approached by these agencies, and E2E-encryption will keep it out of the hands of dragnet surveillance, snooping telecoms providers or FAANG companies and a number of other very realistic scenarios.
(That is unless you use a system that helpfully uploads unencrypted copies of everything you write to a certain large FAANG class company.)
Then there is federation, I guess that is what you mean by decentralized. This is another really sweet property of a messaging system. However, for security it has a number of problems - a little bit less or more depending on exact implementation.
On top of this there is technical execution: Everyones (including me) darling Signal for example has had some pretty nasty problems.
Alltogether it comes down to this if you want to avoid problems: stay out of what trouble you can stay out of and think opsec if you cannot.
And remember that US knew exactly when Soviet started working towards an atomic bomb, because after they started the flow of related research papers stopped ;-)
I.e. opsec is seriously hard if you are up against a powerful entity. Plausible deniability, blending in etc can be just as useful as bulletproof crypto.
I can't speak for GP's use of scare quotes around "decentralized", but it's worth noting in this case that Briar is fully decentralized, not federated. It is peer-to-peer messaging.
This has always been the case for any form of communication, verbal or otherwise. You're always trusting that the people you're talking to don't share your conversation if you don't wish it to be shared. This is a inherent to communication, not anything technology can possibly fix.
If you consider this "the" fundamental problem you're (a) underestimating the severity of other problems and (b) wildly overestimating what technology can achieve for humanity.
Trusting other individuals in this world is not a problem that's solvable by messaging protocols. Even literal scifi solutions like Mission Impossible exploding sunglasses inherently trust the recipient of the message not to share data post-self-destruct. Recipient trust is a fundamentally implicit part of deciding to communicate at all, through any medium.
Services like Signal are great because they're E2EE by default and they're user friendly. We got my tech illiterate mother and stepfather on Signal so they could participate in family group chats. No way she would have been able to navigate GPG.
As for the issue you're bringing up about the endpoint getting compromised, the simple solution is a retention policy (disappearing messages), which Signal has supported for some time now. It doesn't help if the allegiance of owner of that endpoint is flipped (they can simply screenshot future messages), but it does prevent the adversary from getting a full text dump of previous conversations if they swipe the phone for example.
Signal might be a bad example because it is not really that user friendly when it comes to the hard bit. That is: confirming that you are actually connected to who you think you are connected to and not some third party. In a usability study involving Signal[1], 21 out of 28 computer science students failed to establish and maintain a secure end to end encrypted connection.
We should not kid ourselves into thinking that the usability of end to end encrypted messaging has been solved. It is very much still an outstanding issue.
[1] https://www.ndss-symposium.org/wp-content/uploads/2018/03/09...
iMessage isn't an option because half of the family is on Android.
If a motivated state actor wants pictures of my baby niece then I guess they're welcome to it.
I think it is related to forward secrecy.
Maybe the E2EE apps should have a button for that, call it something snazzy like "The 5th Amendment button" and right next to that is a button for trivially forging texts and screenshots with the app's own UI, so that everyone knows that anything can be fake.
But this is just cryptographic repudiation, meaning that a receiver cannot prove that they didn't tamper with their Signal data to fake having received a message from a certain sender. OP is talking about something else: preventing any data the receiver might have received to be transferred to a 3rd-party, but that's impossible.
PGP emails doesn't even have forward secrecy. Emails are not messaging, it needs video/voice calls, stickers/gifs etc etc to have any hope of being adopted by non-techy folks.
The Signal blog has a number of articles on how they develop state-of-the-art privacy preserving features. [2][3][4][5][6].
Also the only info Signal has about you is "Unix timestamps for when each account was created and the date that each account last connected to the Signal service", which is what it provides to government requests [7].
[1] Disappearing messages
https://signal.org/blog/disappearing-by-default/
[2] How to build large-scale end-to-end encrypted group video calls:
https://signal.org/blog/how-to-build-encrypted-group-calls/
[3] Signal and GIFs
https://signal.org/blog/giphy-experiment/
https://signal.org/blog/signal-and-giphy-update/
[4] Signal groups,
https://signal.org/blog/signal-private-group-system/
[5] Sealed sender
https://signal.org/blog/sealed-sender/
[6] Private contact discovery
https://signal.org/blog/private-contact-discovery/
[7] Government requests
Yeah, that is a bit of a mystery. There is no technical reason. I think that email users just want to keep their old emails around, which of course makes forward secrecy pointless. Perhaps PGP users would prefer to use the greater security available for the private key material in an offline medium like email to make it so they don't get compromised in the first place.
Figuring out a solution for this would enable new kinds of horror, like when in 1984 Winston is shown a photo that implicates the party leader in a crime and his torturer then tosses it in the "memory hole" to burn it up. You can see evidence of something horrific but you cannot share it, cannot prove it and then it disappears.
If the counterpart can see something with their own eyes (or a screenshot) then how would you even expect to protect against this?
That's a biology problem, not a digital one.
These systems can't solve that problem, so yeah, you have to trust that person.
But beyond that, good secure messaging protocols are designed so the recipient can verify you sent a message but not prove that to anyone else.
And they also make it so a compromise of your private key can't be used to decrypt old messages.
To expand on my previous post: what's to stop the recipient from taking a photo of the message using another device? Or transcribing it and republishing it manually? Or even just memorising it and telling someone else verbally?
Literally the only defence you have against the recipient abusing their trust is not to send the message to begin with. Everything else is just security theatre.
This is literally the same point others have made too. So why I was downvoted but others not I don't know.
In today's world, offline communication is safer, because to tap it you need manpower ( = more resources, more money ), and that is not efficient for those who want to tap you.