Hacker News Reader

Comment by Animats | Hacker News Reader

Take a close look at the way this works. If there are enough hash matches, Apple apparently gets to look at all the pictures on the phone. The description says "threshold reached, all content visible and unlocked". "All", not "Matched".

Matches can be forced from the outside, by sending pictures to the phone. They don't even have to be from the known-bad image set. Using an adversarial system to try to create hash matches might allow generating innocuous pictures which trigger a phone dump.

Expect a toolset for doing that to be developed. Although it may take a while for word of it to leak out. This has applications for entrapment, swatting, political retribution, etc.

View on HN

Even regardless of whether it's all photos or all matched photos, the issue is that Apple is actively and publicly building adversarial black boxes that run on your device, on a OS level (on iOS AND macs!), with the sole and explicit goal of snitching on you to the government for possessing information that is deemed wrong.

The actual specifics of what material is currently deemed wrong is irrelevant, because it will be expanded.

This. Is. Not Okay.

This is something you'd expect China to mandate, not a Western company like Apple to advertise as a feature.

I don't see why this is surprising at all. Apple has never cared about the privacy of their Chinese users. Privacy was never a principle for them, merely a marketing feature for technically illiterate westerners.

You are a 100% right and I feel so ashamed of myself to have fallen for that.

> If there are enough hash matches, Apple apparently gets to look at all the pictures on the phone.

Not according to the technical summary PDF they released[1], "Only those images that have a voucher that corresponds to a true CSAM match can have their vouchers’ data decrypted [...] Even if the device-generated inner encryption key for the account is reconstructed based on the above process, the image information inside the safety voucher for non-matches is still protected by the outer layer of encryption"

[1] https://www.apple.com/child-safety/pdf/CSAM_Detection_Techni...

This is spin by Apple, driven by ignorance or deliberately misleading statements. NCMEC's database contains images that are not CSAM, not illegal and are not even borderline (grey area).

The fact there is a match in NCMEC's database does not mean the content is CSAM. It does not even mean it's illegal. In fact, it doesn't even mean there's a single person in the picture frame.

And no, this is not theoretical. NCMEC's database is this much of a mess today.

In fact, given there's a threshold at all I'm pushed into believing that Apple know the database is faulty.

> NCMEC's database contains images that are not CSAM, not illegal and are not even borderline (grey area).

Got a handy link about this? A quick search doesn't show anything obvious.

> This is spin by Apple, driven by ignorance or deliberately misleading statements.

The claim was that Apple got access to all the pictures on the phone which isn't correct according to their technical summary. Nothing to do with the validity of NCMEC hashes.

I think the problem is that no one will ever be able to validate what's in that database. "I need to see that database so I can ensure that it's all CP" isn't going to go over well.

Anyone can already do that by sending a iOS user an incriminating message via airdrop or sharing an album. Unless this search is limited to only photos saved to the personal Photos Album, all apps on the iOS device become an evidence plant attack vector including the built in mail app, third party apps that can accept image messages or have planted illegal images, and even iMessage.

This is especially important given that NSO Group's Pegasus' works through sending multimedia iMessages/WhatsApps with exploits. The former auto-uploads to iCloud, and the latter can be configured to auto-upload all received media to iCloud.

It almost sounds like a feature Apple is building to explicitly support the imprisonment of dissents through tje false planting of child pornography.

What the actual f--k?

> It almost sounds like a feature Apple is building to explicitly support the imprisonment of dissents through tje false planting of child pornography.

But you can already do this without these changes - cf the UK police officer currently under investigation for having CP on her phone that someone else sent her and (she claims) never opened.

https://www.theguardian.com/uk-news/2020/aug/02/former-polic...

Once again, "People can already do this" is not a rebuttal or a defense.

It's a rebuttal to "is building to explicitly support the imprisonment of dissents" when this is already possible without this new feature.

I like to think this case will call for reform in these laws.

"Williams was sentenced to 200 hours of unpaid work and placed on the sex offender register for five years, which damages her chances of getting another job.

Williams completed her community service in a charity shop, and having finished the hours the court ordered her to do, returned to volunteer further."

Reply on news.ycombinator.com