Not clicking on that link, don't want my G account go poof
Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn
11–20 of 108 posts
Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn
#12"Now you can block people in Drive. To prevent people from sharing unwanted files with you, ..."
hahaha. What a coincidence, Google!
So you got a hold of the neural hashes, and then used an error function and descent to generate images that match a 'hash'?
It feels wrong to call them 'hashes' when they're so weak to pre-image attacks. They're not the same idea as cryptographic hashes at all.
Also want to underline how spooky it is that some of them do resemble human forms.
Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn
#13These harmless generated images have a neuralhash equivalent to those provided in the NCMEC database submitted for testing. I repeat: Dont upload these harmless images to iCloud as Apple will assume its Child Porn (CSAM). Scripts were available on a GitHub repo but were removed because they may cause damage to others.
Is the hash database and the hashing algo public? Or how do you know these match?
Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn
#14Do you have any proof?
The database of 200_000 images used by Apple (and others?) is private, and I did not found any trace of the hashes (but I could made a mistake here). So, how do you know that those correspond exactly (or with a certain threshold that has NOT been disclaimed by Apple) to the CSAM DB?
Also, NeuralHash has NOT been released by Apple yet (https://www.apple.com/child-safety/pdf/CSAM_Detection_Techni...), so...
Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn
#15I've never used modern apple products, but I have a question on how apple cloud works: is it possible that simply sending these in a messenger software to someone who uses apple cloud for automatic backups could get that person flagged as a child pornographist?
Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn
#16As far as I know they all do CSAM, Google, Microsoft, Facebook, and now Apple with iOS 15. So isn't it already a problem that you have it in a Google Drive?
I put it in Google Drive on purpose to analyze if Google is also scanning as aggressively as Apple. So far no warnings.
Did you try it with Apple, though? Or did anyone else?
Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn
#17These harmless generated images have a neuralhash equivalent to those provided in the NCMEC database submitted for testing. I repeat: Dont upload these harmless images to iCloud as Apple will assume its Child Porn (CSAM). Scripts were available on a GitHub repo but were removed because they may cause damage to others.
> Dont upload these harmless images to iCloud as Apple will assume its Child Porn (CSAM)
This is not true. They may match the hash, but the will not match the visual derivative.
The system is not as easily fooled as you think.
Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn
#18I've never used modern apple products, but I have a question on how apple cloud works: is it possible that simply sending these in a messenger software to someone who uses apple cloud for automatic backups could get that person flagged as a child pornographist?
My understanding is that this is the whole point and one of the main reasons people get so upset about it.
Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn
#19I've never used modern apple products, but I have a question on how apple cloud works: is it possible that simply sending these in a messenger software to someone who uses apple cloud for automatic backups could get that person flagged as a child pornographist?
No, because they won’t actually match. The poster is just wrong.
Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn
#20The problem with Apple’s approach to CSAM is that they use Neuralhash. Unlike other simple perceptual hashes, the failure modes and the collisions using this method are not well understood. I repeat here my previous comment in another thread : they use NN and triplet embedding loss, the exact same techniques used by neural networks for face recognition, so maybe the same shortcomings would apply here. For example a team of researchers found a 'Master Faces' that can bypass over 40% of Facial ID. Now suppose that you have such an image in your photo library, it would generate a ton of false positives and not just a single match with the NCMEC database.