CSAM is Apple’s new neural network which will scan through all of your photos (on Icloud or locally on iPhone) and is used to detect child abuse. CSAM stands for “Child Sexual Abuse Material”.
Here is a demo of how it is supposedly meant to work. This demo is in Python and is only an example and that is probably not what Apple will use.
- Neural networks hash all of your images.
Image 1 which is a picture of clouds (clouds.jpg):
Image 2 is a picture of a house (house.jpg):
2. If these neural networks wanted to compare these clouds to this house, they compare the hashes (00041ebdfffffffb and 000cef3ef0f07800) and see if they match up. This is what happens on your phone, except with your personal pictures and child safety organizations. They also use longer hashes for comparisons.
3. If a match is found (something like 000cef3ef0f07800 and 000cjjuea0f07800), it will be reported to the authorities or to parents.
There is also a problem with this system.
The systems can be fooled easily by applying a hashtag filter. The custom filter sends a custom hash to the servers and the servers just accept it.
And another one.
The police/anyone can send you an image of something harmless, to which they have a filter on and CSAM detects that the image is dangerous and starts sending police to your house.
This system is such a massive problem and has so many fatal flaws and also a LOT of privacy violations and is getting released in 2022.
If you want true privacy, then switch to Android because Apple will make you feel like you are constantly under surveillance (because you are) with their bad systems.
Sources: https://www.forbes.com/sites/gordonkelly/2021/11/11/apple-iphone-ipad-warning-security-privacy-csam-hack/ and https://www.bleepingcomputer.com/news/technology/researchers-show-that-apple-s-csam-scanning-can-be-fooled-easily/
Thanks for reading.