3 ms·
Or manipulating photos a little so the classifier deems them dangerous: https://www.youtube.com/watch?v=YXy6oX1iNoA https://www.youtube.com/watch?v=YXy6oX1iNoA
by jll29 5y ago
Or manipulating photos a little so the classifier deems them dangerous:
https://www.youtube.com/watch?v=YXy6oX1iNoA https://www.youtube.com/watch?v=YXy6oX1iNoA
- isaacimagine 5y agoOr sending manipulated but otherwise innocuous photos to someone else so the classifier is triggered on their device. Yikes.
- judge2020 5y agoThis isn’t a neutral network classifier, it’s an algorithm that searches against existing CSAM photos. https://www.apple.com/child-safety/pdf/CSAM_Detection_Technical_Summary.pdf https://www.apple.com/child-safety/pdf/CSAM_Detection_Techni...
- ChrisMarshallNY 5y agoThat's potentially, even worse. It means that the 'relevant authorities" can substitute any images they want, whenever they want.
- tomas_nort 5y ago“ Messages uses on-device machine learning to analyze image attachments and determine if a photo is sexually explicit.” https://www.apple.com/child-safety/ https://www.apple.com/child-safety/