In theory this is meant to be one of the advantages of end to end encryption: no more "accidental" leakage of user data between users, leakage in logs, etc (remember Facebook's logging incident? [0]). as it's only available on end user devices. And if you look at Apple's documentation [1], they say that iCloud is end to end encrypted. This is obviously not accurate as Apple keeps decryption keys for themselves. But this issue is even worse: here, the end to end encryption was circumvented in such a bad way that this bug could surface.
One thing to remember is that end-to-end encryption comes with the very real possibility of complete and total loss of data, which is hard to stomach for someone’s entire photo library.
iMessage is end-to-end encrypted, but a key to decrypt messages is stored in iCloud backups.
Your statement is nonsensical. Either Apple has the key, or it's end to end encrypted. Both cannot be true. And in fact Apple has the key, even if you turned off iCloud backups in most cases since the person you're talking to probably has it enabled. Which makes the marketing misleading at best, and downright fraudulent at worst.
BTW both Apple and Google have a way of doing end to end encryption of backups with a recovery option in the case of device loss. Apple deliberately chooses not to enable it for iMessage and iCloud photos (it is used for keychain passwords among other things). Credible news reports state that Apple did this at the explicit request of the FBI. https://www.reuters.com/article/us-apple-fbi-icloud-exclusiv...
Also note that Apple policies prohibit anyone else from offering cloud backups for iPhones, so you're SOL if you want cloud backups and privacy. Apple's way or the highway.
> Credible news reports state that Apple did this at the explicit request of the FBI.
A minor point of fact (I have posted this Reuters link more than anyone): a careful reading of the article does not specify an explicit request. It is very much implicit in the article. There is no direct reporting on the request itself.
It is obvious that there was an explicit request. The article simply strongly implies it, because Reuters will not report as fact things they can not verify.
Yes, I apologize for using the incorrect label. iMessage is not end-to-end encrypted for most people.
The point I was trying to make still stands. The recovery options are a form of key escrow. They require a recovery code. As I'm sure you know, if you lose your device and forget the recovery code, then it's total loss – the HSM will reset after a fixed number of tries. People are plenty capable of doing that, and with a billion devices, some non-trivial number of them do it regularly.
Plenty of people lose their devices but I question how many simultaneously forget their screen lock code that they enter literally every day. Of course anything that is possible will eventually happen given enough users. But that doesn't come close to justifying denying the rest of us the option and marketing it as end-to-end encrypted despite that.
Do you have any statement from Apple themselves or any indication that this was their line of thought? As far as I know, the on-device iMessage image scanning for CSAM (presumably what you are referring to) is a shipped feature live on all recent iOS builds, and I don't ever recall seeing any sort of public debate regarding whether iCloud photos should be encrypted or not (to be fair, before this thread, I assumed that they were).
Right, they say iMessage is end-to-end encrypted, but this is only true if you and the people you talk to don't use iCloud backups. If you do, then iMessage's E2E encryption silently degrades to key escrow.
I've read that Apple could technically inject a 3rd key in a two party chat and eavesdrop on it. Not sure how true that is in practice, but it seems plausible based on their security design document.
This is true. It's been proven to have happened in multiple court cases.
They can basically insert another "end" for the end to end encryption without any of the parties knowing.
You basically should never trust Apple / Google / Microsoft / Amazon / etc to handle your private information... ever. Use audited open-source messaging apps.
That article does not say what you claim it does. It’s about feds accessing imessage data via icloud backups, not by injecting keys to tap into conversations.
Using an “audited” app like signal on an iPhone still requires you to trust apple, because they could replace a library signal depends upon or they could just replace the whole app and you wouldn’t know. I also don’t know the extent to which we can verify that the app we get from the App Store is actually the audited version. I’ve always imagined signal could have secret code that gets included at compile time for certain platforms that could make it more vulnerable.
If we controlled all the code on our device and we could build the open source app ourselves that would go a long way. Otherwise you still have no choice but to trust your OS provider.
> You basically should never trust Apple / Google / Microsoft / Amazon / etc to handle your private information... ever. Use audited open-source messaging apps.
Nonsense. Understanding one’s own threat model is critical to deciding the acceptable amount of trust to place in these companies, but black and white thinking helps no one.
So could literally anyone who writes and distributes end-point software that encrypts and/or decrypts content (or hosts software that does) - including operating systems, browsers, e2e messaging apps, password managers, VPN clients, etc.
I believe solving this problem is the crux of the next major breakthrough (if it ever comes) in privacy and personal security. I'm not even sure a solution exists, but a lot can happen when smart people put their heads together on a seemingly intractable problem.
[0]: https://krebsonsecurity.com/2019/03/facebook-stored-hundreds...
[1]: https://support.apple.com/en-us/HT202303