This picture should be presented to everyone who activates iCloud Backup.
“Our end-to-end encrypted system is only really encrypted if you don’t touch our janky unencrypted backup service that we practically beg you to use.”
Maybe if Apple implements some really good automated scanning in iMessage, the government might allow me to encrypt my backups.
“The encryption keys will be provided with the content return” sure sounds bad.
When we wrote securephones.io, we had a theory that iCloud Backup was uploading raw disk encryption keys. This is what the documents say. But it seemed too stupid actually be exploitable.
People keep saying “what’s the big deal if you don’t have anything incriminating on your phone.” People: the issue here is not that the cops can break into your phone with a warrant. It’s that if *they* can do it, so can someone who hacks your account.
Anyway, WhatsApp has announced E2E encrypted backup. Google has announced E2E encrypted backup (though they don’t encrypt enough). Apple is increasingly looking like the outlier on this feature.
It’s not possible to maintain a “privacy-focused” image and yet do nothing on encrypted backup — while your competitors slowly outcompete you on encryption features.
The most optimistic explanation I have for the scanning fiasco this summer is that Apple knows it’s rapidly falling behind, and this was a desperate attempt to negotiate a solution that buys them some flexibility with governments.
Now the problem here (and I will keep predicting this until it comes true or I look dumb), it is very obvious to me that Apple is going to ditch client-side scanning soon and go to server-side scanning for iCloud Photos.
Which leaves end-to-end encryption for iCloud Backup in an unknown state. Will we ever get it, or does Apple just give up on the feature forever because they’ve committed to server-side (plaintext) scanning?
Lots of people say “you can back up your device locally” but seriously, how much longer will that even be supported? I find it hard to believe that my (now teenage) children will have any other option *besides* cloud backup by the time they graduate from college.
Anyway, this idea that “cloud” and “device” are two different things is a shining artifact of the past. In the future they’ll be one seamless whole, and if your private files on one have to be scanned — then there are no private files on the other.
It’s just amazing to me how huge the stakes are just in this one decision, and people think this is just about some backup feature. It makes me want to drink.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
I still don’t understand why “let’s make system RNGs fast and insecure in case someone wants to run a Monte Carlo simulation” was ever considered to be anything other than the dumbest of all possible thoughts.
Apparently there is this huge population of language/OS users running Monte Carlo simulations and THEIR LIVES WILL BE UTTERLY RUINED if those simulations run slow.
I mean if I’m doing statistical simulations and the built-in system PRNG is slow, what’s the impact? I guess I’ll have to spend the 10 minutes copy/pasting the Mersenne Twister off Wikipedia. Seems pretty non-devastating.
I headed over to the Home view to see why tech Twitter has gotten so much lamer, and it’s all people speculating on when Bitcoin will hit 100K.
So henceforth this account will just speculate on the Bitcoin price.
I’ve been using Latest Tweets for a couple months now and switching back to Home is like going home to find your parents have turned your bedroom into a Taco Bell.
Many governments realize they’ve “lost” their chance to ban end-to-end encrypted messaging, but they also realize it doesn’t matter because unencrypted backups are much more useful.
So predictably the vanguard of the conversation has shifted away from E2EE (which is in many places a done deal) to device backups — which are still not widely E2EE encrypted.
I personally think that governments want to preserve unencrypted backups because it provides investigative capability. (Ie. they can subpoena Apple to get your texts.) And CSAM fear is just a tool that politicians are using to preserve this capability. But 🤷♂️.
Good thread about the recent Unicode attacks and some previous work that predates it. I agree that citations could be improved. But I want to push back a little. 1/
What was interesting to me about the recent Unicode/Trojan attacks (link below) isn’t that Unicode contains some exploitable fluffery. *Of course* it does. Unicode is terrible. 2/ trojansource.codes
What was surprising to me is how many compilers, source management tools and IDEs were vulnerable to the attacks. I expected this from pomo languages like, say, Golang or Swift. But even compilers for ancient languages like C/C++ were happy to eat Unicode and not complain. 3/
Fortunately I’m clever and I’ve checked my Dropbox into Github.
I keep every academic project since 2003 in a directory named src2/. Why src2? Because six years and three laptops ago I somehow corrupted src/ and was afraid to overwrite it. In 2025 I anticipate an upgrade to src3/.