News for hackers. Stuff that matters.

A comment on Apple defends anti-child abuse imagery tech after claims of ‘hash collisions’.

Re:Apple defends anti-child abuse imagery tech after claims ... by zakember on Wednesday August 18, 2021 @04:52PM (#28226976)
Is no one going to talk about how Apple is implementing this?

If Apple is training a neural network to detect this kind of imagery, I would imagine there to be thousands, if not millions of child pornography images on Apple's servers that are being used by their own engineers to train this system

[ Parent | on HN ]
Re:Apple defends anti-child abuse imagery tech after claims ... by arn on Wednesday August 18, 2021 @04:53PM (#28226991)
It's not a neural network. It's not AI. It's not trying to interpret image content. It's trying to identify known specific images, but is using a fuzzy fingerprinting match.
[ Parent | on HN ]
Re:Apple defends anti-child abuse imagery tech after claims ... by zimpenfish on Wednesday August 18, 2021 @05:11PM (#28227225)
> If Apple is training a neural network to detect this kind of imagery

NCMEC generate the hashes using their CSAM corpus.

[ Parent | on HN ]