They also appear to have killed the crazy part of scanning on device. That’s the crazy part, not scanning in iCloud.
Google is not just scanning for known examples of kiddie porn, they are scanning for images of naked children, and reporting parents to the police.
for instance, they tried to get a parent arrested for using their Android phone to get medical assistance for their child during lockdown.
>With help from the photos, the doctor diagnosed the issue and prescribed antibiotics, which quickly cleared it up. But the episode left Mark with a much larger problem, one that would cost him more than a decade of contacts, emails and photos, and make him the target of a police investigation.
“I knew that these companies were watching and that privacy is not what we would hope it to be,” Mark said. “But I haven’t done anything wrong.”
The police agreed. Google did not.
https://www.nytimes.com/2022/08/21/technology/google-surveil...
Windows defender may even be checking the files it scans for whatever content MS considers objectionable. We likely wouldn't won't know unless it becomes a news story or a whistleblower tells us.
Google has apparently shared their algorithm with others, but only Facebook was mentioned by name.
>in 2018, Google developed an artificially intelligent tool that could recognize never-before-seen exploitative images of children. That meant finding not just known images of abused children but images of unknown victims who could potentially be rescued by the authorities. Google made its technology available to other companies, including Facebook.
https://www.nytimes.com/2022/08/21/technology/google-surveil...
Instead I think the average user is just not aware of the insane amount of data that is exfiltrated and what can be done with that data, because the software does not make it clear what it actually is doing.
Even more cynically, it just gives them more data about you and your interests (with viewing frequency/duration being a measure of interest) which lets them push more narrowly targeted ads at you, which gives them more money, which lets them invest more in improving Windows as a whole or whatever else they tell themselves so they can face their own reflections.
* https://www.reddit.com/r/Windows10/comments/8zk1yy/a_simple_...
It stores that information in a local index, and also stores telemetry data which counts interactions with the app, but there is no evidence it is uploading indexes to MS.
Do you have any evidence of this?
https://learn.microsoft.com/en-us/windows/configure/windows-...
https://learn.microsoft.com/en-us/windows/configure/basic-le...
They both 404 now. I had saved a copy of both at the time, but those files must be on another drive. I'll keep looking though!
There were several articles about it at the time https://www.extremetech.com/computing/247311-microsoft-final...
https://www.ghacks.net/2017/04/06/windows-10-full-and-basic-...
These documents were massive and filled with lots of information buried in places that seemed odd and scattered across different sections so that you couldn't get all the information about what an app collected in one section/table of the document.
Another thing I remember from those pages was that in some circumstances short clips (less than 1 minute) of the videos you played using window's default player would be sent to MS as well, but they said that would only happen in the event of crashes.
But that's not the point. Victim blaming doesn't do any good for today's victim, but maybe it will help show the next person that their fate is under their control.
Google is trying to get people arrested because Android backed up their personal photo libraries to Google Photos.
> The day after Mark’s troubles started, the same scenario was playing out in Texas. A toddler in Houston had an infection in his “intimal parts,” wrote his father in an online post that I stumbled upon while reporting out Mark’s story. At the pediatrician’s request, Cassio, who also asked to be identified only by his first name, used an Android to take photos, which were backed up automatically to Google Photos. He then sent them to his wife via Google’s chat service.
https://www.nytimes.com/2022/08/21/technology/google-surveil...
> Not scanning for CSAM on your own servers isn't a realistic expectation.
Google isn't just scanning for known examples of kiddie porn, they have developed an algorithm that scans for any image of naked children and, as usual, there doesn't seem to be a human in the loop when their algorithm makes a bad call.
Not having a human in the loop is terrible and a problem frequently when it comes to getting support from Google companies. That's a separate issue from Google or Dropbox being able to scan unencrypted files for CSAM. Google's policy of automating as much as it can has tons of downsides. But it's understandable when you look at the scale Google functions.
It's important to separate the policy of scanning from Google's terrible appeal process and the algorithm false positives.
I would feel differently if the story was that Android scanned an outgoing SMS or a photo saved locally. I am not sure where the balance point is to identify and report CSAM while also respecting user's rights to privacy.
The slippery slope claim made in the past was that they were only looking for files that match known examples of kiddie porn.
For instance, this tech from Microsoft:
>PhotoDNA creates a unique digital signature (known as a “hash”) of an image which is then compared against signatures (hashes) of other photos to find copies of the same image. When matched with a database containing hashes of previously identified illegal images, PhotoDNA is an incredible tool to help detect, disrupt and report the distribution of child exploitation material.
https://www.microsoft.com/en-us/photodna
Reporting parents to police because they took pictures of their child's first bath and those photos were automatically backed up to Google servers is much, much worse.
1. The picture(s) in question mentioned above which got an account shut down and police notified was not a 'child in a bath', it was close ups of a child's genitals
2. You are probably not a parent since a child's 'first bath' is not a thing. They are bathed as infants starting immediately. If you are referring to a child's first independent bath, then no one should be taking pictures since private bathing is not something that should be intruded upon for picture taking
All that said, I agree with you.
Are you a parent? A prudish parent perhaps?
I have pictures of me as a kid in a bath (genitals obscured underwater). It's a common parental thing among parents who AREN'T DRIED UP PRUDES AND FRIGHTED OF NUDITY.
The bottom line is - another case of our ability to use our technology as we desire being interrupted and interfered with. I hate to even remotely invoke RMS, but... He wasn't wrong!
Please try to parse a response before yelling at someone for something they didn't state.
No, the example of Google calling the police on a parent given in the NYT article was because the parent sought medical assistance for their child during lockdown and the doctor requested pictures of the issue the child was having.
This is no more an acceptable reason for Google to report parents to the police for child abuse than taking pictures of your child's first bath, and taking pictures of your child's milestones has long been something that people have done.
> No, the example of Google calling the police on a parent given in the NYT article was because the parent sought medical assistance for their child during lockdown and the doctor requested pictures of the issue the child was having.
These things are not mutually exclusive.
Stop being defensive. It doesn't help you. I am trying to help you by telling you that you are making a shitty argument.
OK. Allow me to tell you that, in my opinion, you are the person making the shitty argument
Your argument that "no one should be taking pictures" of their child's first bath is just cringe-worthy.
Google is to blame for the situation when they call the police over a single false positive from their faulty algorithm, not parents who take pictures of their own children.
You are obviously either unable or unwilling to parse what I am trying to tell you, and fixated on this notion that you have to be 'right' no matter what even though I am not telling you that you are wrong!
I am at a loss as to what to reply to you with besides I urge you to read things before continuing to respond because otherwise you end up talking around people instead of with them.
Good day.
The whole photos experience on Android has been simplified to break that distinction and there is no way to even really know or realize what is happening - BY DESIGN.
The real answer here is, again, Android is not privacy first by default.
BTW parents taking medical pictures of their children != CSAM which is an acronym for "Child Sexual Abuse Material"... Taking care of your child's medical problem isn't "sexual abuse" despite the conflation of the two people are glossing over in this thread.
There's a very clear distinction between local and cloud, every single photo taken using any camera app is local only. Galleries, like Google Photos, and backup services like Dropbox have an explicit setting to enable cloud backup. Google Photo backup is very distinctly different from Google Drive and phone backups in general.
I use FolderSync to sync it with a self hosted NextCloud instance. If I am travelling, I prefer using Syncthing, a wifi file server running off the phone, or a physical wire to transfer photos. I know exactly what pictures are sent where and backed up with what method. I am not sure why you believe Android treats photo and file management like a black box, I don't believe Apple does either. Apple iOS devices are much more difficult to manually backup as they only allow background photo sync to iCloud, if the screen is turned off syncing to a third party will most likely fail. But there is a straightforward toggle to disable Photos backup to iCloud.
Android file management is superior to any other mobile OS. Android is file privacy first by design, as long as you are comfortable managing backup and syncing with self-hosted and self-managed solutions. I have gone the additional step of rooting my phone as well to get around any Android limitations on what folders certain apps have access to.
No one is arguing against the last point you made. No one was conflating the two.
>Android is file privacy first by design
Yet we have the NYTimes article stating that the only ways Google could have access to the photos they tried to have a parent arrested for would be to scan content on the device, scan automatic backups, or spy on the text messages sent from the device.
The examples in the story were parents sharing images directly with their doctor, not the with the general public.
But you could maybe argue that if the ML model is based on CSAM then it's a product of it and would need to be banned.
>Ed Francis studies the evolution of military technology over at his YouTube channel, Armoured Archives. But this week, Francis says five years’ worth of research stored on Google Drive has become inaccessible thanks to Google’s automated error.
Francis says the file in question was simply a collection of data on various tanks for a coming video on how military vehicles have evolved across historical conflicts. But Google’s automated systems deemed the file a terrorist threat, resulting in a complete lockdown of his YouTube, GMail, and Google Drive accounts.
https://www.vice.com/en/article/qj8yj7/google-locks-historia...
Having a shitty algorithm kill your whole Google account with no way to reach a human to fix the problem is one thing.
Having a shitty algorithm report you to the police for taking pictures of your child's first bath is a bridge too far.
What stood out to me what one of the first couple comments in the support thread was the contact information for the Museum Director of the Swedish Tank Museum.
My unconfirmed theory is that Google's OCR pdf service flagged specific text and pages in the pdf of tank plans and repair manuals as they are considered classified and shouldn't be in the public unredacted. It's historically significant and absolutely worth preserving, but I can see why it might get flagged.
This was also interesting in that it got the entire account taken down. Usually Google flags a file and disables sharing completely. Google disabled sharing backups of a recent Kanye interview that was filled with hate speech called "Kanye champs removed video.mp4". I have not watched it but it was my first time seeing something removed because it "may violate Google Drive's Hate Speech Policy. Some features related to this file were restricted."
I am not sure if it was the post to HN or various videos other made that helped, but it made me realize that the one of the best ways to get in touch with a human at Google is posting here on HN and that Google will continue to become more aggressive with scanning files.
Section 230 was modified by FOSTA-SESTA when related to hosting sexual solicitation, and EARN IT is back under review by congress which would effectively make E2EE illegal and make hosts liable for illegal content uploaded by its users.
Would scanning be okay if there is a government entity with a human you can appeal to that would override any flags made by automated systems? No corporation wants more government oversight, and the only way to avoid it is to do good faith self policing. EARN IT would receive a lot more support if the public sentiment changes to believe Google and Apple and other hosts intentionally choose to ignore problematic materials. Reddit is criticized often because it allowed problematic subreddits to grow. I am not sure where things will end up.
* Can't remember who said it but it was at a town hall this year
This just sounds like yet another excuse for holding payroll down as much as possible.
If I am a customer of Amazon, Apple, Netflix, Walmart or any number of the other companies with a similar market cap, I can get access to real live human beings who provide customer support.
I don't get why that should be the case. When you rent an apartment the owner is generally no longer able to just walk in and do whatever he likes, including installing cameras in the shower. So I would expect that when Google tries to sign you up for its online storage (repeatedly) you should get the same protections, especially with smart phones being such an important part of modern life.
From one perspective, I understand that cloud storage should include certain rights to privacy. If I buy some compute time on AWS, I manage and control all flow of data. It would be a complete violation if AWS policed the kind of content I could store. I think self-managed vs company managed is what dictates my expectations. Google photos and drive, especially since most people don't pay for it, fall into the company managed category in my opinion.
In the last few years Google Drive became more strict about sharing copyright work and started flagging files, which limits the ability the "share to anyone with a link". For people who use it as a backup, the advice I started hearing was to encrypt the data before uploading. The opsec while considering nation-state level threats to privacy already recommended only using encrypted containers since the beginnings of cloud storage.
The Communications Decency Act is an interesting law that dictates the liability of online content and service providers. There is an ongoing battle to increase ISP and host liability as a way for the entertainment industry to combat piracy.
FOSTA-SESTA and the ruling on Backpage [1] made things more difficult for hosts like Google and changed the liability surrounding sexual solicitation. If a host like Google knows what's being hosted and does nothing, they get in trouble. They don't have the protections they did in the past.
Again, the parents Google is reporting to the police are not trying to share photos with anyone except their own doctor.
This isn't scanning information because you made it publicly available. This is scanning information on your device because it was automatically backed up by Android.
Just to give context to perspectives people might have on this analogy, the above would be illegal in e.g. Germany.
It is if the false positives are overwhelming, given their impact on innocent people's lives.
What makes it not a realistic expectation? According to other references, the USA cannot compel companies to run scans on their own customers.
I own my device, I don't own their cloud. That's a big difference. Don't co-opt my property to do work you want done. Data stored on your servers is your business, so doing the checking there is fine, as long using them isn't mandatory.
Are you trolling? What Federighi proposed before was scanning "for CSAM" on device [1]. Same angle.
> Doing it on device is probably preferable to on cloud and at worst no different.
Please elaborate. How is it better to force users to run software they don't want than to let them decide whether or not to have their photos scanned when they choose to upload them to the cloud?
Anyway it's a false dichotomy. Apple isn't doing on-device scanning, and now they've announced they won't do it in the cloud either.
So yeah allegedly no worst.
You might notice i used the word "allegedly" a lot, it's because we are speculating about a feature that was never actually rolled out and that nobody audited externally to my knowledge. If you don't trust Apple then this argument don't apply and you are probably better of not using an iPhone.
Nonetheless it's still not worst than actually rolled out CSAM scanning feature of Google Cloud that already had major user adversial effect. So you should trust Google even less and you definitely shouldn't use a stock Android device.
What was crazy about this? The media outrage around the PR was incoherent to me at the time, and it seemed no one took any time to understand the present realities or details.
What’s crazy about this?
Another way of expressing the concern: does your iPhone work for you, with the help of Apple’s services, or does your iPhone work for Apple? Working for me means not having software designed to report me to the police for how I use my device. The hash database for now includes CSAM hashes; there’s no reason it couldn’t be extended to include similarly heinous material, like Winnie the Pooh memes, Hong Kong freedom posters, rainbow flags, hormone therapy instructions, or anything else offensive to the regime.
People that are comfortable with their device working for a company with the assistance of the user choose Android.
It wasn't like constantly monitoring your private residence, more like setting up a breathalyzer checkpoint on the way to the freeway.
Your apartment complex decides they don’t want to be party to anything illegal. Just in case, they set up a police precinct in the lobby. They set up hidden cameras in the lobby, the hallway leading to your apartment, and the balcony of your apartment. "All public spaces of course" the crowd consents. And if their AI model detects anything suspicious, they send the video to the detective. Because you aren’t doing anything illegal, you have nothing to worry about, right?
Everyone scans for CSAM. I am not conjecturing on the ethics of scanning photos here, I am suggesting that moving from server- to client-side scanning had no effect on any of the things you are ranting about. Hence why I do not understand the outrage.
> Working for me means not having software designed to report me to the police for how I use my device.
Let me fix it: “how I use iCloud Photos, a hosted service on Apple’s servers”.
You agree _in your rental contract_ not to dump explosives in the apartment complex trash compactor, and that the apartment complex has the right to process your trash.
The apartment complex stops looking inside everyone's bags that land in the compactor (a bit late to prevent it from getting into the trash), instead sniffing bags as you put them into the trash chute on your floor, and tagging the bag with a red sticker if it the chemical composition matches a particular pre-registered explosive from a list.
At the other end of the chute, a counter checks if you've dumped a dozen bags with red stickers, and if so, a human may open one to see WTF you're up to. If you're indeed dumping explosives in bulk, they let the bomb squad know.
EITHER WAY, that bag would be headed to the compactor, you are affirmatively putting it there, and any explosives are a breach of agreement. In no other case besides you opening the chute to dump a bag is it sniffing anything else anywhere, except when you open the chute to send these bags to the compactor, bags you decide the contents of.
Once the feature exists your local data might become accessible to a government warrant, which would make the iPhone the opposite of a privacy oriented device.
If it's only for iCloud uploaded data they can simply do the scanning there. There's no reason to use customer's CPU/battery against them.
This is specifically what Apple is trying to avoid - they are intentionally pushing an environment where the user must be involved to get the keys, by way of their account password and/or other enforcement mechanisms designed to ensure only the real user can access such keys.
This is discussed in detail in their Apple Platform Security guide: https://help.apple.com/pdf/security/en_US/apple-platform-sec...
All of the facial recognition, object identification, etc is all done on-device for the same reason. By contrast Google can and does do this in the cloud - and there are Google servers that intentionally have access to decrypt your photos.
iCloud backup was previously a vector that bypassed this, however they also today announced they are fixing that: https://news.ycombinator.com/item?id=33897793 "Apple introduces end-to-end encryption for backups"
But there are many serious problems with it. I'll isolate one: an on-device content surveillance mechanism is a slippery slope to a bad, bad place.
As the saying from the 90s goes, "child porn [CSAM] is the root password to the US Constitution." It is bad enough that people are willing to suspend their better judgement to do something about it.
But after you already have the mechanism accepted an in place, it is far easier to add to it. Pick your boogeyman: you're giving all of them a lovely tool to address their desires. Dissent suppression and Winnie-the-Poo detection for Xi, Erdogan gets to sniff out Golem memes, and choose-your-own horror for the coming dictator of the US.
It is far harder to tell a sovereign, "we could easily do that, but will not" than "we don't have a mechanism to do that." And pretending that it won't happen doesn't pass the laugh test - we have seen this show many, many times. But if you want to argue, start by explaining how Apple's jumping to implement the 10-minute-max sharing limit shows how they'd stand up to China about this.
It would help to note Apple also today announced "Advanced Data Protection" in iCloud which closes the hole where iCloud Backups, iCloud Photos and various other bits of data were technically decryptable by Apple. They've closed that (but it's opt-in, to balance the average user losing all their photos against other users desire to be secure even if it means losing their data). Details: https://support.apple.com/en-us/HT202303#advanced
However even without the "Advanced Data Protection" what I said about not having a workflow for any Apple server to "normally" request both the keys and photo data is also still good security.
This is what Apple was trying to avoid. Scanning on iCloud also requires that Apple can see your photos.
If the scanning is done on device, Apple could encrypt the photos in the cloud so that they can't decrypt them at all. Neither could the authorities.
> There's no reason to use customer's CPU/battery against them.
The amount of processing ALL phones do for phones is crazy, adding a simple checksum calculation in the workflow does fuck-all to your battery life. Item detection alone is orders of magnitude more CPU/battery heavy.
Why does that dystopia require on device scanning? Why couldn't they just do it with an OS update today? It's not a reasonably slippery slope, given the actual mechanics of how the CSAM system was designed (perceptual image hashes, not access to arbitrary files)
> There's no reason to use customer's CPU/battery against them.
That's the better argument, but still not super strong. On-device scanning means you know and can verify what hashes are being scanned for, who is providing them, and when they change. Cloud scanning is a complete black box. None of us would know if Google was doing ad-hoc scans of particular users' photos at the behest of random LEOs.
See my comment here:
https://news.ycombinator.com/item?id=33903825
> None of us would know if Google was doing ad-hoc scans of particular users' photos at the behest of random LEOs.
Not your device, not your software. You should assume anything you upload unencrypted is scanned. This distinction was clearly voiced by the majority during the debacle of Apple's on-device scanning proposal. They basically said, "Scanning in the cloud is [choose one: fine, skeezy], but we draw the line at doing on-device scans. I don't want that software on my device."
That would be precisely as difficult to verify as verifying whether your Apple device is currently scanning your content.
Apple can run software that represents Apple's interests on their own servers.
This was not some Google-style hotdog or not algorithm that was using "AI" to guess if a photo had a nude person on it or not.
Second, the overall aim of the technology also targets legitimate or inadvertent activities, that the law has also unjustly criminalized - say taking family pictures of your kids, 17 year olds sexting, or someone trolling you. Once again unjustly branding people as "child abusers", this time with "evidence" to back up the narrative.
But even in the case of a proper match to the NCMEC list for its bona fide purpose, that still is undermining the phone's owner's interests. That's the philosophical contour, even if you personally wish to brush it aside with the desire to catch people looking at evil images. Our society respects similar longstanding privacy boundaries, even if it does end up helping some "bad people".
It's really not in the NCMEC's interest to respect any of this, as their foundational dynamic is that there is a horrible thing happening in the world, and they must do everything possible to stop it. That kind of advocacy is certainly needed, but we shouldn't just accept their desires as if they're an unbiased neutral party.
Everyone just forgot that in the Apple system you don't get automatically banned and reported to the authorities if you get a match (or five). The matches get manually checked by an actual human, who will see in 0.2 seconds that it's not actual CP, but a highly distorted picture of a cat or some gibberish. Zero action will be taken.
Also: the scanning was only done if iCloud is enabled. If iCloud is enabled, the EXACT SAME scan could be done in the cloud. Why are people not objecting to this with the same fervour?
2. What's going on with the "fuzzy comparison"?
[0] https://en.wikipedia.org/wiki/National_Center_for_Missing_%2...
Which is weird, because it implies anything you used but didn’t buy can and should betray you, and it’s your own fault. You should have been smarter. But it makes a bit of sense, for some odd reason I can’t quite make tangible. When the samsung TV I bought starts showing me ads, when my pricey cable or Hulu package interrupts for commercials, I feel a tinge of this.
So now I think I somewhat get the feeling. Thank you.
The contrapositive isn't automatically true. You can be upset when a cloud host or ISP violates your expectations, even though you don't own them. You can expect Apple to basically work in your interests even on their own servers, since you're paying them.
The ownership thing is more about the structure of society. If computing and communications technology is going to move us forward as a distributed democratic society, then those capabilities must remain distributed throughout the population rather than being controlled by a handful of centralized gatekeepers.
And yes, we've strayed quite far from this ideal. Most people have little control over what their mobile phone does. Still, there is a distinction between a device not having your desired functionality because the manufacturer didn't create it (or even disallowed others from creating it), and the manufacturer deliberately creating functionality that harms users.
Running it on their servers means nobody knows how expansive the scans are, and there is nothing preventing ad-hoc scans for arbitrary content for individual users.
The actual scanning and apple self-interest all happened server side.
And to accomplish that all we have to put up with is a huge lack of privacy and tons of false positives?
Now the authorities can go "there might be CP in there, think of the children!" and a judge will give them the warrant.
On-device check agains KNOWN CP checksums + E2EE encrypted iCloud -> authorities need to get the actual phone, they can't just go looking around on people's iCloud accounts to see if they might find something.
Are you arguing in favor of on-device scanning? Or arguing against all scanning?
Perhaps easier, what do you think they should be doing?
Either commit and say “scanning for CSAM is the top priority” and scan on device or say “privacy is the top priority” and don’t scan at all.
I’m literally saying this halfway path is the worst of both worlds, since people still lose privacy, but very little meaningful progress against csam will be made.
No, it’s definitely still there.
One program scans photos that were about to be messaged and if they look like nudity asks "are you sure you want to do that" (and also optionally notified the parents, although it looks like they aren't going through with that part). The scanning happens on device, and the results of the scanning are never sent to Apple or any other third party.
The other program scanned all images that were being uploaded to iCloud and if it found any that matched a government maintained image fingerprint database would notify Apple who would forward that information to the government. Big difference in scope and impact.
Communication Safety for Messages is opt-in and analyzes image attachments users send and receive on their devices to determine whether a photo contains nudity ... The company told WIRED that while it is not ready to announce a specific timeline for expanding its Communication Safety features, the company is working on adding the ability to detect nudity in videos sent through Messages when the protection is enabled.
.... “Additionally, because the minor is typically sending newly or recently created images, it is unlikely that such images would be detected by other technology, such as Photo DNA. While the vast majority of online CSAM is created by someone in the victim’s circle of trust, which may not be captured by the type of scanning mentioned, combatting the online sexual abuse and exploitation of children requires technology companies to innovate and create new tools. Scanning for CSAM before the material is sent by a child’s device is one of these such tools and can help limit the scope of the problem.”
Yes, I am being very cynical here for 2 major reasons - (1) Governments around the world want control and access to our device, and BigTech can deliver. This is not an Apple or Google problem, this is a social and democratic problem that we need to fight politically with the government by demanding stringent data protection and privacy laws. (2) Apple is a corporate who needs to be profitable. Collecting data and selling it (either to the government through programs like PRISM or to advertisers through a platform is a very lucrative source of revenue. Anybody who things Apple will let go of billions of dollars from that is delusional (and that is exactly why Apple is pivoting to become a service company). Apple's history when it comes to invading its user's privacy is just as bad as Google's or Facebook.
And to make things more secure, the only way apple could read those fingerprints was if you uploaded 5 csam photos.
How is that worse than Google where engineers can look at every photo you upload regardless of the content as part of their server side scanning?
Only if you exclude the following network transmission, which is the easy part that needs no special code. The privacy concern comes in with those two things together. So yeah if you take away one half of a bad thing such that the bad thing no longer becomes possible, it's not bad anymore. The concern is the whole process, on-device scanning being the key not-yet-implemented component.
> If Apple has a legal obligation to ensure that iCloud does not store CSAM/etc. then either you have to scan on device before upload _or_ you have to store iCloud data without E2E encryption.
Apple does not have that legal obligation. If they can't decrypt the content on their servers, then their only response to a government-issued warrant would be to hand over encrypted data.
Also, CSAM is not the concern. The concern is this would be used against dissidents in authoritarian countries. On-device scanning takes us a step towards becoming one and further empowering the existing ones.
CSAM is a serious problem and you're ignoring Apple's moral or even repetitional obligation to try to address it. However poorly.
We shouldn't be hyperbolic and just flatten levels of badness, or we'll can never find a compromise (which is, I suspect, just how you want things).
My understanding is in the USA companies like Apple cannot be legally obligated to ensure that iCloud does not store CSAM. Something about the US Constitution, but I can't remember what. Apple is legally obligated to report CSAM if they come across it themselves though.
This appears to be the case in Europe as well. But may not always be that way. The EU appears to be working on legislation that can compel cloud providers to scan for CSAM: https://9to5mac.com/2022/05/11/apples-csam-troubles-may-be-b...
Welcome sources on this from others. Last time I dug into this was a year ago.
The first amendment comes into play. The government cannot compel Apple to write software in a certain way, such as "write your encryption so you have keys that access all of users' data". That would be "compelled speech". So if the government provides Apple with a warrant, Apple can only provide encrypted or whatever meta information they have, not decrypted content.
I disagree. On-device scanning is desirable here and scanning it in iCloud though the industry-norm, is not desirable.
Apple specifically stated they would only scan the photos being uploaded to iCloud [1] and not local photos which were not being uploaded.
Most cloud storage providers are scanning in the cloud, this necessitates a workflow where both the photos and decryption keys are accessible by the same server and that there is a security workflow to request the users decryption keys without the user involved (by way of their account password, or other enforcement mechanisms designed to ensure only the real user can access such keys, Apple has previously detailed Hardware HSM enforced details of that for iCloud in their Apple Platform Security guide [2]).
By extension that opens up the potential for a malicious employee or third-party to either intercept those photos while being scanned or otherwise allows a workflow that could be exploited to request the keys and photos. This is a major loss of security and privacy which Apple is specifically and intentionally designing to avoid. Until now iCloud Backups were a major weakness in that story however they also today announced they are fixing that [3].
Hence: the on-device scanning was just an entirely technical solution to scanning "only photos uploaded to icloud" (like everyone else) while also "ensuring some server in icloud can't decrypt your photos".
You could of course argue that it makes a later change of policy to scan local-only photos easier. While this is true, Apple could decide to add code to iOS to do anything they want at any time, so they could just decide to do that anyway, even if they were scanning iCloud photos in the cloud. At the same time, by not scanning them in the cloud, they have a technical solution to ensure it's much harder for a malicious actor to get unauthorised access to iCloud user photos. Because there is no place/server that is "supposed" to be able to access both the photos and the keys.
[1] https://www.apple.com/child-safety/pdf/Expanded_Protections_... [2] https://help.apple.com/pdf/security/en_US/apple-platform-sec... [3] https://news.ycombinator.com/item?id=33897793 "Apple introduces end-to-end encryption for backups"
What are the benefits? It would tell Apple (and hence law enforcement) who had copies of existing, known CSAM. I guess that would help catch such people. But such individuals would also have strong incentive to simply turn of iCloud to completely avoid detection.
What are the risks? Apple employees rummaging through your photos (after 30 NeuraHash matches, a human reviewer would go in and check your photos, and as some analysts pointed out, hash matches can be spurious, that is, completely different photos can have the same hash by coincidence).
Over time, would the program expand beyond CSAM? Consider the intersection of countries in which Apple sells iPhones and countries that criminalise being LGBTQ [1]
[1] https://www.humandignitytrust.org/lgbt-the-law/map-of-crimin...
However my argument is specfically to the OPs point that they should just scan the photos "in the cloud" and not "on your device". But the reality is they were only scanning photos that were being uploaded to the cloud, they were just doing it on your device so that Apple doesn't need to have a server that can decrypt your photos to do it in the cloud. Which is what most others (including Google) do.
I think you might have misinterpreted whatever those analysts wrote. The chances of thirty images all coincidentally matching both perceptual hashes is essentially nil.
The human reviewer would've gotten "reduced quality images" of the matches and would make a decision based on those.
And yes, people could've just turned off iCloud. That's how encryption works.
I can’t see how that is worse than unencrypted photos in the cloud that can be scanned at any time server side.
That the server may know the hash is IMHO much less a problem (it shouldn't be reversible anyway) than the legitimization of on-device scanning.
Let's say tomorrow China tells Apple to do client-side scanning, and let's say Apple decides to comply.
Scenario A (Apple already did on-device scanning):
Apple flips a few bits. iOS no longer cares whether images where uploaded on not. The code change can easily be hidden assuming no reverse engineering. There's little measurable difference unless the user has a matching image in which case the user is already done for [EDIT: in retrospect, the device should be running a scan on the local images every blacklist fetch. I still think that scanning more images is much less of a difference than the other scenario, where the device has to do stuff it just did not do before]
Scenario B (Hash Comparison is server-side):
The code change can be hidden of course. But otherwise there are significant changes:
The device needs to calculate hashes to every image upon creation which it did not before. The device needs to fetch the blacklist every X days which it did not before. The device needs to run the local scan every blacklist fetch which can be slow if many images.
All these things are far more likely to be noticeable and far more likely to generate network traffic.
>Pushing a perceptual hash of every photo to Apple would actually decrease your privacy as now Apple can find any known photo in your library - not just CSAM but on any topic - memes, political protests, porn, whatever.
The Hash shouldn't be reversible, so the only way for Apple to do that is to have an existing dataset. Which is possible, but preferable to the no E2E situation and (IMHO) the on-device scanning situation.
This seems far better than having every photo scanned server side and have those photos sit there unencrypted.
Of course that's a lot less secure and leaks more privacy info, but the objections seem more about the morality of having a device you own doing the scanning, not so much about any privacy concerns.
Lots to look at there but the key bit is:
>If the user image hash matches the entry in the known CSAM hash list, then the NeuralHash of the user image exactly transforms to the blinded hash if it went through the series of transformations done at database setup time. Based on this property, the server will be able to use the cryptographic header (derived from the NeuralHash) and using the server-side secret, can compute the derived encryption key and successfully decrypt the associated payload data
So my understanding is that the server won't even know the hash of the image unless the client-side operations determine that it matches one of the hashes in the CSAM database, which is stored on the client.
Are you seeing it differently?
My own devices should not snitch on me to the government. Obviously this mandatory snooping of all private data is not going to stop at the single most politically salable application; that's just the starting point.
The point is, I think if Apple decided to scan your photos for CSAM even if you weren't using iCloud, that this would be a distinctly different situation to scanning only the ones uploaded to iCloud.