It would scan photos you were uploading to iCloud, not private photo libraries on your phone. I'm sure you'll agree it's important to correct such a misunderstanding as one of those is a lot more invasive than the other.
Since late 2022 Apple has enabled Advanced Data Protection, which encrypts all photos before they’re uploaded to cloud storage. With ADP on, my photo library is “private” not just in the common-language sense (it contains extremely personal data I have not shared with others) but also in an opinionated technical sense that these files are accessible only to me. If Apple’s CSAM scanner was deployed today, it would be scanning those photos in cleartext on your phone before unreadable data was sent up to the cloud. You could argue that Apple was making a trade: “hey, it doesn’t matter whether we can read the private data you’re storing, the price of sending even unreadable encrypted private data to our infrastructure is that you must run local software that scans the private photos on your phone,” and that’s a trade you might accept or reject on the merits. However I think it’s extremely important to say it exactly this way and not play language games. Apple was going to mandate local scanning of private photos as the cost of using their infrastructure even to store opaque private bits.
It did a great job at freaking people out hearing about their photos getting scanned and it could be defeated by making a 1 pixel change to any photos a pedo would hide on their phone (since any changes to the image would totally change the hash).
This isn't the way those hashes work. A 1 pixel change would still hash similar enough to be matched. Maybe there are adversarial 1 pixel changes that could break the hashing, but I doubt it.
Even cropping, watermarks and other manipulations like that would still match. "Perceptual hashing", very different to cryptographic hashing. It's basically checking if an image looks "similar enough".
I believe this is why they needed multiple matches, because otherwise there must have been too many false positives.
This may be too oversimplified, but imagine that in a series of CSAM images, there might be, for example, a wall or furniture or something, that could appear similar enough to a wall in one of your own photos. That's a match, off to the gulag with you!
> "If Apple’s CSAM scanner was deployed today, it would be scanning those photos in cleartext on your phone before unreadable data was sent up to the cloud."
Or enabling Advanced Data Protection could have disabled the scanner, we don't know. Even if it went the way you said, you could still not use iCloud and have private photos on your device, whereas your phrasing is trying to imply that there would be no option to do that.
I think we will have to agree to disagree about the idea that turning on cloud backup suddenly makes my private photo library “not a private photo library.”
[1] https://www.reuters.com/article/us-apple-fbi-icloud-exclusiv...
Not capable of what, running software? Communicating with a HTTPS endpoint? Having library code? Running stuff in the manufacturer's interest rather than your interest? All those things happen already in some form or other, and there isn't a cutoff to make the phone incapable of it without hobbling the phone.
> "That's one big step closer to the described full-private scanning (and just a flag flip away)."
iPhone already does scan offline private photos for face and object recognition purposes. And run big blobs of unknown Apple-provided code. It's only your trust in Apple that makes you think it doesn't report anything back now - and nothing at all stopping them from being arm twisted by the authorities to make that scan for something the government dislikes and report on it, as you say a flag flip away. It already does send your location and your surrounding WiFi signals and your voice when you use Siri unless you toggle the privacy settings, and that all came in quietly on regular updates.
Apple walked a fairly narrow line when they announced it, and when they publicly stated that if the authorities asked them to extend the scope of the scanning that they would refuse.
I don't know why they chose to do it on the endpoints rather than in the cloud, but acting like doing it on the cloud would give you any level of protection from them putting intrusive software on your phone is not reality. (Same with Google, Samsung, et al).
Not loaded with trained models on illegal content and wired up to alert the authorities if it finds a match, with presumably several teams within Apple built around that feature. I'm thinking about more than the technical aspects of this.
> It's only your trust in Apple that makes you think it doesn't report anything back now
Yeah, exactly. I trust them enough right now to run tons of stuff without my knowledge on my phone. I don't have the time or knowledge to audit my phone, even if it were Android. If they announced that new feature is going live like it's a thing customers are meant to be ok with, I'd trust them a lot less.
So they move the scanning to the device.
It honestly seems to me like they thought they could negotiate a middle ground without pissing off the Feds or the customers, but they maneuvered it quite badly.
Oh cool, didn't know that. It's this new "advanced data protection" feature that makes everything in iCloud e2ee except the classic mail/contacts/calendars combo that wouldn't really work with that. https://support.apple.com/en-us/HT202303 is a nice resource on this, and I wish more companies would publish things like this.
I'm a bit too chickenshit to try it, as losing my devices is all too likely, but I'm glad it's available for those who need it.
There was no way to separate the private photos from iCloud-uploaded photos. It was all-or-nothing, like Android permissions: “Allow govt to scan all your private pictures, or do you wish to have no backup?”
I was perfectly feasible to design the ability to have private photos, but Apple chose not to. Or Apple, in collaboration with the government, chose not to.
(Yes it was feasible to design the ability to have a local photo store which isn't uploaded to iCloud, separate from other photos which are, and call it "private photos", but that's another matter).
It's a slippery slope that ends up with your phone/computer snooping on texts, call contents, or anything else and then submitting your "crimes" to the authorities.
> "Such an expansion would use the exact same justification as the iCloud-bound content scanning."
One of the justifications was that Apple are/could be legally responsible for criminal images hosted on their servers, that exact same justification wouldn't apply to offline content scanning.
> "It's a slippery slope that ends up with your phone/computer snooping on texts, call contents, or anything else and then submitting your "crimes" to the authorities."
Not "crimes" in quotes, crimes without quotes. Generally people think law enforcement is important, especially regarding crimes against children. "A slippery slope" which leads criminals being punished is not the argument winning logical fallacy you think it is. The argument against it is around invasion of privacy, rights not to self-incriminate or to remain silent, ownership of device and software, freedom from unreasonable search without prior evidence, whether you can be found guilty by algorithm, et al.
You're missing my point. Content scanning will start with the Four Horsemen of the Infopocalypse[0]. Then it'll move on to "crimes" like blogging about a public figure[1].
[0] https://en.wikipedia.org/wiki/Four_Horsemen_of_the_Infocalyp... [1] https://www.flsenate.gov/Session/Bill/2023/1316/BillText/Fil...
> The media erroneously reported [a statement from September 2021] as Apple reversing course.
Linking to an article from December 22, 2022 in which Apple is quoted as literally saying “We have further decided to not move forward with our previously proposed CSAM detection tool for iCloud Photos.”, ie actually and literally reversing course.
If this person cannot comprehend an article, should I trust that they actually checked that what they were seeing was what they were paranoid about?
Btw, the screenshot shows an interaction with api.smoot.apple.com, which is known to be used for spotlight and related services: https://news.ycombinator.com/item?id=8479958 ; mediaanalysisd is used for visual lookup, where macOS tries to identify landmarks and other items in an image to help you find more information about it (https://appleinsider.com/articles/23/01/21/tests-confirm-mac...).
In short, paranoid people are good to use as indicators for further investigation, but are rarely to be trusted as sole sources—even less so when they attempt to ascribe intent.
Your note about that domain is interesting though. I don't regularly use Apple devices, so I'm not particularly concerned by that anyways.
Yes there can be private conversations between two people, but there can't be private conversations between two people where one of them isn't trusted but can hear the conversation and simultaneously, what, can't hear it to keep it private?