Hacker News new | ask | show | jobs
by xoa 1752 days ago
>All of which has been common for a long time

No, this has never been done before.

>You where already sending your unencrypted phots to Apple

No, I wasn't. The whole point here is that Apple is not scanning server side, they've built the functionality to scan device side. You need never use iCloud in any way whatsoever in order for Apple's new scanning tech to be used against you. That is a major difference.

>Apple and all other service providers are required by US and most other countries laws to do searches on their servers.

No, they are not, even if that was the new thing Apple is doing which it isn't. If a company builds server-side scanning, then they may be required to fulfill certain requirements. But companies are not required to actually do that in the first place even if many choose to do so. Apple already did scan uploaded photos and voluntarily chose not to have E2EE for iCloud data in order to please law enforcement agencies, but that's a voluntary choice by Apple. This new client side scanning is a different beast. Please try to gain even the slightest fucking clue what you are talking about before spouting off on something so important.

Edit: to add for those interested in more details on the law, the federal reporting requirements are under 18 U.S. Code § 2258A [0]. What you'll see there is a "Duty To Report", and the reason for that is to evade Constitutional protections. If the government compelled companies to scan, as well as any legal challenges (by very well funded actors) and public blow back, as a practical matter those companies would become State Actors for the purposes of 4th Amendment evaluation. However, even if it's heavily incentivized so long as it's "voluntary" courts have repeated ruled that 3rd parties can do searches that would be illegal for the government, turn discovered evidence over to the government who in turn may then use it freely. Walter v. United States (1980, [1]) is a good example, covering the [righteous and just prosecution] of someone transporting "films depicting homosexual activities" after it was mailed to the wrong address and turned over to the FBI which I'm sure everyone here on HN would applaud and definitely is what they have in mind when they think of client side scanning in the US. Tim Cook is carrying on that tradition with Pride no doubt.

----

0: https://www.law.cornell.edu/uscode/text/18/2258A

1: https://www.oyez.org/cases/1979/79-67

1 comments

> The whole point here is that Apple is not scanning server side

False. Apple complies with the laws pertaining to customer data and provides data as legally required. III. Information Available from Apple

J. iCloud

https://www.apple.com/legal/privacy/law-enforcement-guidelin...

> You need never use iCloud in any way whatsoever in order for Apple's new scanning tech to be used against you.

False. Phones don’t download the CSAM hashes so they can’t do device side scanning as they have nothing to compare the images to. Yes, the phone uploads a hash, but they also upload the unencrypted images along side it.

Thus the only thing that changes is Apple isn’t paying for the compute power to do the hashing. That and a tiny amount of extra bandwidth on uploading images.

PS: In response to your edit, perceptual hashes are a grey area. However, as long as a judge agrees they can very much just take down production systems when it pertains to a case. That’s a rather big stick to force compliance even if it’s not an explicit law it’s very much a consequence of it. Thus companies really don’t push back as some that have simply got raided.

>> The whole point here is that Apple is not scanning server side

>False.

No, TRUE. Apple announced TWO difference technologies. The first one is specifically for completely independent machine learning client-side scanning of all messages. From Apple's own announcement:

>First, new communication tools will enable parents to play a more informed role in helping their children navigate communication online. The Messages app will use on-device machine learning to warn about sensitive content

Which has been clarified to mean any and all sexually explicit material, and then notifies the parents. Apple is billing this as only for child accounts and only to parents, but that is merely a set of flags and directions in the programming not anything inherent to the system. It could be applied to any ML model at all and the notifications sent to anyone at all. The system is now built and ready for governments to compel Apple to use for other things in complete violation of device owner's rights, backed by Apple's total ownership of device root.

The second feature is the one for client-side scanning of all photos uploaded to iCloud for illicit content using "neuralhash" which is subject to collisions as have already been repeatedly generated and received lots of discussion here as well as elsewhere (ie., [0]). That is claimed to be initially aimed at uploads and CSAM only, not that there is any way to be sure, but again same thing: the system now exists to perform arbitrary fuzzy scans on-device whether someone is uploading elsewhere or not.

This is absolutely new and horrible capability. If you upload something unencrypted to somebody else's property, as well as the law there is a reasonable common sense understanding that you're depending upon their good will and that they may be compelled at that point without having to involve you. "Possession is 9/10 of ownership" and all that, regardless of details. Now one's own personal private property will have built-in locked down systems to scan all your data based on arbitrary third party choices.

----

0: https://news.ycombinator.com/item?id=28219068

> "It could be applied to any ML model at all and the notifications sent to anyone at all"

This is the kind of FUD I keep calling out. Your position "it doesn't matter what they actually built, because I can fantasise a completely different system". Do you not see a massive gaping problem with that reasoning - to whit, it could apply to literally everything with the same predictive power? iTunes music match could look for any media, therefore it does. iOS update could install malware, therefore it does. Smartphone microphones could be listening, therefore they are. iOS and Android could send your GPS location back to head office even if you flick the software switch telling it not to, therefore they do. Apple pushed a U2 album onto your device, therefore they will push pro-China MP3 lectures onto your device and you will be forced to hear them.

It's bad (low quality) reasoning done in bad faith.

> "The system is now built and ready for governments to compel Apple to use for other things in complete violation of device owner's rights, backed by Apple's total ownership of device root."

Which they have said they won't do. Of course "facts don't matter, my fantasies are more real" is your position again.

What an laugh of a trash post. There are plenty of angles of ignorance and outright ludicrous statements here, but your finale really underlines it:

>Which they have said they won't do.

What the hell? Apple just won't comply with governments using force against them? How in your magic world do you imagine them doing this? If China says "you will add this specified model to your detection or else we will kill iPhone production" what exactly do you imagine Apple is going to do? Apple has already caved there, repeatedly, such as with iCloud data itself. Or if the USA points out, correctly, that while it can't compel Apple to create something from scratch it is on much stronger ground to compel Apple to use capability it already has and successfully wins that now what? They're just going to defy the US Government and the feds will just shrug and say "oh nevermind jodrellblank already said on HN Apple won't do that"?

You throw around "bad faith", "FUD", and "fantasies are more real" and then just outright accept Apple saying "they won't do" something which equates to defying government. This is not going to strike anyone actually grounded in reality as a particularly "fact based" bit of reasoning on your part.

> "say "oh nevermind jodrellblank already said on HN Apple won't do that""

I didn't say Apple won't do that, I said "Apple said they won't do that". I'm not making a claim of what will happen, I'm reporting a fact of what Apple said, which was this:

"“We have faced demands to build and deploy government-mandated changes that degrade the privacy of users before, and have steadfastly refused those demands. We will continue to refuse them in the future.”" and "“Let us be clear, this technology is limited to detecting CSAM stored in iCloud and we will not accede to any government’s request to expand it.”"[1]

[1] https://www.cnbc.com/2021/08/09/apple-will-reject-demands-to...

> that is merely a set of flags and directions in the programming not anything inherent to the system.

Parental controls can also be setup to track kids real time position, such tracking is invasive but that’s what the people buying the phones want. It’s a much larger stretch to say Apple is going to suddenly enable that and track every user in real time for example even though they 100% could.

Anyway, it’s all just code, every device with over the air updates has the exact capacity you fear. As it stands no it doesn’t do what you’re worried about. It could after someone builds as system to do more than just detect porn, but that was true of the original iPhone.

So, I get the concern but if that’s what you’re worried about you should be equally concerned for every other device that can be updated without your consent.