3 ms·

China can make that regardless of what Apple thinks, that’s totally unrelated to this. Also this technology does not develop similar images, the image has to b

• gzer0 5y ago
  This is incorrect. The images in this link [1], are completely different. One is a cat, one is a dog. Same hash produced for both. [1] https://github.com/anishathalye/neural-hash-collider https://github.com/anishathalye/neural-hash-collider
  • djrogers 5y ago
    That’s a hash collision on the first of two distinct perceptual hashing algorithms that are used before a human review. It’s mind-numbingly unlikely that 2 images would match 2 unique perceptual hashes and a human review and not be the same image.
  • robertoandred 5y ago
    Do you have an example using a version of NeuralHash from less than 9 months ago?
• coolspot 5y ago
  Previously there was no precendent of an on-device content scanning protocol proposed by the top phone manufacturer themselves and even deemed privacy-preserving and secure by them. Now please idividually hash every word of every message on the phone. Then give us all illegal messages that contain _exact_ hash match with the dangerous word “Tianamen”. Government abuse is impossible though: our team of Apple employees here at iCloud-China will manually review them before passing the list to firing squads. Thanks!
  • hypothesis 5y ago
    > idividually hash every word It will likely be a fuzzy match on though crimes. But every citizen will get their budget of 30 or so “security” vouchers, so we can all vent a bit once in a while. Oh and they will also double check it with “totally independent” blacklist on server before running it by a party person last.