Of all the invasive things modern tech does, I'm genuinely surprised it is the child porn scanning that has the media in an uproar.
The media doth protest too much methinks? As yet another journalist or elite gets busted for child porn.
The media doth protest too much methinks? As yet another journalist or elite gets busted for child porn.
Do you invite the police into your home to watch what you do to make sure you aren't breaking any laws?
Here is the relevant paragraph from Apple’s documentation:
“as an additional safeguard, the visual derivatives themselves are matched to the known CSAM database by a second, independent perceptual hash. This independent hash is chosen to reject the unlikely possi- bility that the match threshold was exceeded due to non-CSAM images that were ad- versarially perturbed to cause false NeuralHash matches against the on-device en- crypted CSAM database. If the CSAM finding is confirmed by this independent hash, the visual derivatives are provided to Apple human reviewers for final confirmation.”
https://www.apple.com/child-safety/pdf/Security_Threat_Model...
I had read through the technical whitepaper [1], which does not include this information. Thank you for sharing. Since the second hash only works on pictures that Apple can decrypt within this system ("for an account that exceeded the match threshold"), this merely saves the human reviewers at Apple time.
[1] https://www.apple.com/child-safety/pdf/CSAM_Detection_Techni...
Apple’s CSAM mechanism can’t be ‘picked’ in the way that you claim. The two ‘locks’ are not separate. You can’t pick them one at a time.
> I had read through the technical whitepaper [1], which does not include this information. Thank you for sharing. Since the second hash only works on pictures that Apple can decrypt within this system ("for an account that exceeded the match threshold"), this merely saves the human reviewers at Apple time.
This is false. The second hash is independent and designed to prevent attacks based on adversarial spoofing of neuralhash.
Nobody, not even Apple, denies that someone can generate neuralhash collisions.
You can only assert that this is a vulnerability by making the false claim that a neuralhash collision alone will cause the system to flag an image.
> The second hash is independent
lol
The locks are not separate. You have to open both with the same ‘key’, I.e. the image.
From the documentation:
“the visual derivatives themselves are matched to the known CSAM database by a second, independent perceptual hash”
The “lock” in this case involves one key, and two hashes. If they were separate, you’d be able to ‘pick’ them one at a time with separate keys.
You can’t.
If such a system should be built they surely have gone out of their way to preemptively secure it.
I'm still very much of the opinion that we need people both in police, politics and in tech with balls - or whatever the equivalent is for women - to tell people that we cannot just interfere with everyones right just to hopefully maybe catch a few more baddies, but again:
If such a system needs to be built, this is an amazingly well thought out system.