Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’
31–40 of 465 posts
Re: Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’
#32Earlier quoted context omitted.
They did, they lobbied heavily against EARN IT, and so did groups like the ACLU and EFF. However, it didn't stop Congress members from moving it along through the process - they were only stopped because Congress just ran out of time. It was clear that Congress was not interested in listening at all to the lobbyists on the issue. It's clear that EARN IT could literally be revived any day if Apple didn't do something…
It's clear that EARN IT could literally be revived any day if Apple didn't do something to say "we don't need it because we've already satisfied your requirements." Alternatively, "Apple has shown that it's possible to do without undue hardship, so we should make everyone else do it too".
They were going to legally mandate that everything be scanned through methods less private than the ones Apple has developed here, through EARN IT and potentially LAED (which would have banned E2E in all circumstances and any device that could not be unlocked by the manufacturer). While that crisis was temporarily averted, the risk of it coming back was and is very real.
Apple decided to get ahead of it with a better solution, even though that solution is still bad. It's a lesser evil to prevent the return of something worse.
Re: Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’
#33For everyone upset about Apple's CSAM scanning, I think we all forgot about the EARN IT Act. It was nearly passed last year but Congress was finished before it could be voted on. It had Bipartisan support and would've virtually banned E2E of any kind. And it would have required scanning everywhere according to the recommendations of a 19-member board of NGOs and unelected experts. The reason for this mandatory backdo…
You can't ban encryption, it's practically impossible, it's like banning math.
Its amazing since this would have decimated the American Tech sector in many unknown ways.
Re: Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’
#34Earlier quoted context omitted.
Cool project. Wouldn't padding all generated images with a few items that match CLIP get around this though?
Doing that would mean the NeuralHash would change though. And you'd have to not only get CLIP to identify CSAM in the generated image but also negate the parts of the generated image that are causing CLIP to label it as "generated" (while still colliding with the target NeuralHash). Unclear how hard this would actually be in practice (if I were going to attempt it, the first thing I'd try is to evolve a colliding ima…
Swing and a miss. Not in the CSAM dataset. Take two images. Encode alternating pixels. Decode to get the original image back. Convert to different encodings print to PDF or Postscript. Encode as base64 representations of the image file...
Who are we trying to fool here? This is kiddie stuff.
This is more about trying to implant scanning capabilities on client devices.
Re: Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’
#35> The system relies on a database of hashes—cryptographic representations of images—of known CSAM photos provided by National Center for Missing & Exploited Children (NCMEC) and other child protection organizations. “Cryptographic representations of images”. That’s not the case though right? These are “neuralhashes” afaik which are nowhere close to cryptographic hashes but rather locality sensitive hashes which is a…
Something to note here is that in the hash collision that was discovered, the two images look nothing alike. One is a picture of a dog, the other is blobby grey static.
Re: Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’
#36For everyone upset about Apple's CSAM scanning, I think we all forgot about the EARN IT Act. It was nearly passed last year but Congress was finished before it could be voted on. It had Bipartisan support and would've virtually banned E2E of any kind. And it would have required scanning everywhere according to the recommendations of a 19-member board of NGOs and unelected experts. The reason for this mandatory backdo…
CSAM scanning is only one type of scanning which is performed at the behest of governments and corporations around the world. This alone will not allow for end-to-end encryption of items in the cloud. There would need to be end-device scanning for arbitrary objects, including full text search for strings including 'Taiwan', 'Tiananmen Square', '09 F9', and so forth to even begin looking at e2e encryption of your item…
Re: Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’
#37> The system relies on a database of hashes—cryptographic representations of images—of known CSAM photos provided by National Center for Missing & Exploited Children (NCMEC) and other child protection organizations. “Cryptographic representations of images”. That’s not the case though right? These are “neuralhashes” afaik which are nowhere close to cryptographic hashes but rather locality sensitive hashes which is a…
No - the reporting is absolutely terrible here. 1) These are more share similar visual features than crypto hashes. 2) HN posters have been claiming that apple reviewing flagged photos is a felony -> because HN commentators are claiming flagged photos are somehow "known" CASM - this is also likely totally false. The images may not be CASM and the idea that a moderation queue results in felony charges is near ridiculo…
Not going to happen. Lawyers in the US have issues with offering unsolicited advice, and other problems with issuing advice into states where they are not admitted. So likely none of the US lawyers (and the great many more law students) here will ever put their real name to a comment.
Re: Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’
#38>"This independent hash is chosen to reject the unlikely possibility that the match threshold was exceeded due to non-CSAM images that were adversarially perturbed to cause false NeuralHash matches against the on-device encrypted CSAM database," …
Re: Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’
#39Earlier quoted context omitted.
Why can’t we have the alternative we already have, where we don’t have our phones scans and E2EE exists without compromise? Why do we have to accept it any other way? If you’re trying to frame this as “we need to prevent Congress from ever passing something like the EARN IT act”, I agree. Apple and other tech companies already lobby Congress. Why aren’t they lobbying for encryption?
They did, they lobbied heavily against EARN IT, and so did groups like the ACLU and EFF. However, it didn't stop Congress members from moving it along through the process - they were only stopped because Congress just ran out of time. It was clear that Congress was not interested in listening at all to the lobbyists on the issue. It's clear that EARN IT could literally be revived any day if Apple didn't do something…
That tech is not being deployed on iMessage which is the only e2ee(ish) service from Apple (with Keychain) and is what those legislative attempts are usually targeting. One could argue it would have made sense (technically) there though, sure.
Was it a reason to release it preventively, on something unrelated, to be in the good graces of legislators ? I'm not sure it's a good calculation, and it doesn't cover other platforms like Signal and Telegram that would still be seen as a problem by those legislators and require them to legislate anyway.
Re: Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’
#40> The system relies on a database of hashes—cryptographic representations of images—of known CSAM photos provided by National Center for Missing & Exploited Children (NCMEC) and other child protection organizations. “Cryptographic representations of images”. That’s not the case though right? These are “neuralhashes” afaik which are nowhere close to cryptographic hashes but rather locality sensitive hashes which is a…
They use “private set intersection” (https://en.wikipedia.org/wiki/Private_set_intersection) to compute a value that itself doesn’t say whether an image is in the forbidden list, yet when combined with sufficiently many other such values can be used to do that.
They also encrypt the “NeuralHash and a visual derivative” on iCloud in such a way that Apple can only decrypt that if they got sufficiently many matching images (using https://en.wikipedia.org/wiki/Secret_sharing)
(For details and, possibly, corrections on my interpretation, see Apple’s technical summary at https://www.apple.com/child-safety/pdf/CSAM_Detection_Techni... and https://www.apple.com/child-safety/pdf/Apple_PSI_System_Secu...)