Does anyone know about a good write up of this discussion in Japanese?
Apple’s new abuse prevention system: an antritust/competition point of view
181–190 of 318 posts
Re: Apple’s new abuse prevention system: an antritust/competition point of view
#182Earlier quoted context omitted.
> Problem is that this scanning is necessarily fuzzy and there is going to be a false positive rate to it. And the way that you'll find out that you've tripped a false positive is that the SWAT team will knock your door down and kill your dog (at a minimum). Not true. Hash matches are to be human reviewed. So no, people won't get "swatted" accidentally as you allege. The other concerns people have been voicing are ce…
>Hash matches are to be human reviewed. Until it proves too expensive, then a different AI system will do it instead. I have zero faith that it'll be a fully competent, well trained, well rested, well paid person will actually be doing these reviews in the long run.
I have to imagine that you’ll be investigated and can let Apple rehash your photos and find the ones that tripped the system. If Apple rehashes all your photos and suddenly none of them are tripping, and they accuse you of deleting the bad ones (assuming you haven’t deleted anything) the problem isn’t the hash algo, it’s something else in the system.
That said, I don’t like my phone being a snitch.
Re: Apple’s new abuse prevention system: an antritust/competition point of view
#183Can I just say that I'm fascinated that this is happening under Tim Apple - an openly gay man in his 60s?
I mean this is someone who would know what being on the wrong side of law means - not only federally, but probably quite intimately since Alabama was not exactly known for its acceptance of homosexual people.
Now just imagine if the US still had anti-homosexuality laws (like the majority of the world still does) and your phone is constantly scanning your photos to check for signs of homosexual behaviour. Forgetfully take a selfie with your boyfriend, it gets flagged, sent to the Ministry of Morality, and next thing you know you're being dragged into a van. Best case scenario you're being jail. Worst you're being thrown off a building, stoned, or brutally beaten to death.
That's the future Apple is signing us up for. There is zero chance this stops at CSAM, especially with the Democrats convinced that half the country are the absolute worst people on the planet and not being shy about completely ignoring the rule of law and Constitution to extend the reach of the state to levels that would make a totalitarian blush. This will end terribly.
Re: Apple’s new abuse prevention system: an antritust/competition point of view
#184Earlier quoted context omitted.
Could you point to the regulation that requires this?
See the “Federal CSAM Law” section here: https://cyberlaw.stanford.edu/blog/2020/01/earn-it-act-how-b... > Section 2258A of the law imposes duties on online service providers, such as Facebook and Tumblr and Dropbox and Gmail. The law mandates that providers must report CSAM when they discover it on their services, and then preserve what they’ve reported (because it’s evidence of a crime).
The law doesn't require them to actively go around scanning for it.
Re: Apple’s new abuse prevention system: an antritust/competition point of view
#185Earlier quoted context omitted.
> No. You can design a system where the FPR is essentially zero. Even if shitty md5 is used. How can you do that, considering md5 can have collisions?
A confusing thing that I think people haven't addressed clearly for the most part: For a hash (whether cryptographic or perceptual), there is a chance of random collisions and also a difficulty factor for adversarially-created intentional collisions. The random collision probability has to be estimated based on some model of the input and output space (with cryptographic hash functions, you would usually model them a…
Thanks for the detailed explanation. I understand why that works for perceptual hashes if you make them really precise, however I doubt it would work with md5, which is why I asked.
Re: Apple’s new abuse prevention system: an antritust/competition point of view
#186Can I just say that I'm fascinated that this is happening under Tim Apple - an openly gay man in his 60s? I mean this is someone who would know what being on the wrong side of law means - not only federally, but probably quite intimately since Alabama was not exactly known for its acceptance of homosexual people. Now just imagine if the US still had anti-homosexuality laws (like the majority of the world still does)…
Alternatively, and just as likely, is that the Republicans do the same evil things that you say the Democrats would do. Misusing tech isn't just for the "other" side, my friend.
Re: Apple’s new abuse prevention system: an antritust/competition point of view
#187Earlier quoted context omitted.
No. You can design a system where the FPR is essentially zero. Even if shitty md5 is used.
Perceptual hashes are not exact hashes, otherwise they would be useless for this task; you would just mirror the image or change 1 pixel. They are instead fuzzy classifiers, and thus have non-zero error rates.
How does this relate to the allegation I was responding to? There is a difference between false positives and false negatives you know. False negative is where the criminal evades the system through measures you describe. False positive — what I was responding to — is where Apple incorrectly accuses an innocent person and thus violates their privacy during the subsequent review.
Re: Apple’s new abuse prevention system: an antritust/competition point of view
#188Can I just say that I'm fascinated that this is happening under Tim Apple - an openly gay man in his 60s? I mean this is someone who would know what being on the wrong side of law means - not only federally, but probably quite intimately since Alabama was not exactly known for its acceptance of homosexual people. Now just imagine if the US still had anti-homosexuality laws (like the majority of the world still does)…
> …your phone is constantly scanning your photos to check for signs of homosexual behaviour. Forgetfully take a selfie with your boyfriend, it gets flagged…
Except that’s not what this is doing.
For all intents and purposes, “scanning” is just hashing content before it’s uploaded and comparing that hash to a known database. So, unless your picture had been hashed and flagged before (And if you just took it how could it have been?) then there’s nothing for it to match.
Don’t just conjure FUD from a contrived worst-case scenario.
Re: Apple’s new abuse prevention system: an antritust/competition point of view
#189Apple will release this in their desktop pcs. The major consequence of this is that, a year down the line, microsoft will also be compelled/forced to join this "coalition of the good". After all, they already scan all your files for viruses, it would be a shame if they didn't scan for anything that is deemed incriminating and also call the cops on you. Of course, the children are being used as a trojan horse again. W…
>We don't even mention the giant logical leap from finding someone possessing CP to automatically considering them a child molester, yet someone posessing ~~an action~~ a murder video is not considered a murderer.
That is because the consumption of child porn induces people to create it. The consumption of action movies does not induce people to murder anyone.
Re: Apple’s new abuse prevention system: an antritust/competition point of view
#190Earlier quoted context omitted.
A confusing thing that I think people haven't addressed clearly for the most part: For a hash (whether cryptographic or perceptual), there is a chance of random collisions and also a difficulty factor for adversarially-created intentional collisions. The random collision probability has to be estimated based on some model of the input and output space (with cryptographic hash functions, you would usually model them a…
Thanks for the detailed explanation. I understand why that works for perceptual hashes if you make them really precise, however I doubt it would work with md5, which is why I asked.
The discussion I thought we were having was about false positives and not adversarially induced false positives. For the former the random collisions have a probability of 1/(2^64).
To mitigate adversarial false positives one idea is to use the combination of a cryptographically strong hash along with a randomly selected perturbation of the file. Prior to hashing, perturb the file and submit both the hash and the selected perturbation to apple. Apple selects the DB based on the perturbation and proceeds with matching and thresholding.