Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

drive.google.com

Post reply on HN
Print view View original

51–60 of 108 posts

Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

#51

Post by nitrogen »

zepto wrote: post #44
Spooky23 wrote: post #32

Earlier quoted context omitted.

The thing that everyone has their panties in a bunch about here and a that like an antivirus scanner, there will be a hash match to child abuse images when you send it. The current practice is that Apple, Google, Microsoft, etc scan the content of your cloud storage. The scenario that you described is a risk and has been since cloud providers started scanning 10-15 years ago. Some large companies scan their file serv…

Yes, except that in Apple’s implementation there also a ‘visual derivative’, which is essentially a blurred thumbnail. Both must match to cause a positive. These images may match the neuralshash, although we have no proof of that at all. They will not also match the visual derivative. This whole post is based on incomplete information.

They will not also match the visual derivative.

How can you be certain, and what prevents a generated image from matching both?

Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

#52

Post by dragonwriter »

Applejinx wrote: post #28
ReactiveJelly wrote: post #12

"Now you can block people in Drive. To prevent people from sharing unwanted files with you, ..." hahaha. What a coincidence, Google! So you got a hold of the neural hashes, and then used an error function and descent to generate images that match a 'hash'? It feels wrong to call them 'hashes' when they're so weak to pre-image attacks. They're not the same idea as cryptographic hashes at all. Also want to underline ho…

I assumed that these are reverse engineered from legitimately illegal and problematic porn of known origin. Not sure exactly how you'd go about doing it, but it seems like there might be a process for 'evening out' areas into solid color that maintains the hash? In which case you're running extensive image processing on illegal images and making variations from those very images. More info on how this is done?

> I assumed that these are reverse engineered from legitimately illegal and problematic porn of known origin

I would assume these were engineered by getting the perceptual hash valies, using distance from the hash values in the DB as an error function, and starting with an innocuous image and hash value, and iterating to a collision for each.

Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

#53

Post by Retr0spectrum »

victor871129 wrote: post #50
LucidLynx wrote: post #14

Do you have any proof? The database of 200_000 images used by Apple (and others?) is private, and I did not found any trace of the hashes (but I could made a mistake here). So, how do you know that those correspond exactly (or with a certain threshold that has NOT been disclaimed by Apple) to the CSAM DB? Also, NeuralHash has NOT been released by Apple yet ( https://www.apple.com/child-safety/pdf/CSAM_Detection_Techn…

Im going to release 5 pieces of proof in the next 5 days

This is a lame attempt at trolling.

Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

#54

Post by newscracker »

Is this obtained from a set of images leaked from NCMEC and regenerated to match the expectations of the as yet (and probably forever) unknown Apple’s NeuralHash algorithm as well as the threshold used to flag content for internal human review for a system that’s going to be operational for U.S. Apple device users only when iOS 15 is released?

On what basis is a set of forest-like and post-alien-invasion and post-apocalyptic abstract art is going to get flagged (my poor eyes see one or two that could have some symbolism)?

Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

#55

Post by Retr0id »

LucidLynx wrote: post #14

Do you have any proof? The database of 200_000 images used by Apple (and others?) is private, and I did not found any trace of the hashes (but I could made a mistake here). So, how do you know that those correspond exactly (or with a certain threshold that has NOT been disclaimed by Apple) to the CSAM DB? Also, NeuralHash has NOT been released by Apple yet ( https://www.apple.com/child-safety/pdf/CSAM_Detection_Techn…

The NeuralHash code is apparently included in the latest beta: https://twitter.com/KhaosT/status/1424205967122571268

Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

#56

Post by zepto »

nitrogen wrote: post #51
zepto wrote: post #44

Earlier quoted context omitted.

Yes, except that in Apple’s implementation there also a ‘visual derivative’, which is essentially a blurred thumbnail. Both must match to cause a positive. These images may match the neuralshash, although we have no proof of that at all. They will not also match the visual derivative. This whole post is based on incomplete information.

They will not also match the visual derivative. How can you be certain, and what prevents a generated image from matching both?

I can be certain because I have looked at the images, and they are obviously not CSAM. Since the visual derivative is generated from CSAM, any spoof must look like it could be mistaken at a glance for CSAM.

What prevents a generated image from matching both is that the attacker would need to know what the image they are trying to spoof looks like, in order to make a false positive of both. I.e. the attacker would need a copy of the original CSAM, and the spoofed file would end up looking like it could be at least plausibly mistaken for that exact image.

Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

#57

Post by zepto »

dragonwriter wrote: post #52
Applejinx wrote: post #28

Earlier quoted context omitted.

I assumed that these are reverse engineered from legitimately illegal and problematic porn of known origin. Not sure exactly how you'd go about doing it, but it seems like there might be a process for 'evening out' areas into solid color that maintains the hash? In which case you're running extensive image processing on illegal images and making variations from those very images. More info on how this is done?

> I assumed that these are reverse engineered from legitimately illegal and problematic porn of known origin I would assume these were engineered by getting the perceptual hash valies, using distance from the hash values in the DB as an error function, and starting with an innocuous image and hash value, and iterating to a collision for each.

That technique can’t fool Apple’s algorithm.

For what it’s worth, the null hypothesis is that they are just fakes and the commenter is at best trying to illustrate a point.

Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

#58

Post by zepto »

victor871129 wrote: post #50
LucidLynx wrote: post #14

Do you have any proof? The database of 200_000 images used by Apple (and others?) is private, and I did not found any trace of the hashes (but I could made a mistake here). So, how do you know that those correspond exactly (or with a certain threshold that has NOT been disclaimed by Apple) to the CSAM DB? Also, NeuralHash has NOT been released by Apple yet ( https://www.apple.com/child-safety/pdf/CSAM_Detection_Techn…

Im going to release 5 pieces of proof in the next 5 days

This comment discredits you.

Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

#59

Post by podric »

jsnell wrote: post #22
ugjka wrote: post #11

Not clicking on that link, don't want my G account go poof

Yeah, seems like anyone clicking through is playing with fire. If the description is correct, posting the link is highly irresponsible. (It seems like the right thing to do would be to serve the content from a server the OP controls themselves.)

Maybe I'm an idiot but my curiosity got the best of me and I clicked the link. The photos just look like abstract modern art, although their perceptual hash may match that of known CSAM, I doubt that anyone who clicked the link will get into legal trouble even if Google flags that folder by detecting a perceptual hash match, as they will likely use real people to verify before taking legal action.

Google has been recently focused on cultivating the image that they care about user privacy. The last thing they want to do is call the cops on a bunch of HN users for looking at some abstract swirly pics.

Re: Tell HN: Dont upload these images to iCloud as Apple will assume its Child Porn

#60

Post by threatofrain »

LucidLynx wrote: post #38
sharken wrote: post #27

Earlier quoted context omitted.

> The system is not as easily fooled as you think. I would like to believe that is true, but the negative consequences of even generating a false-positive is enough to not attempt to upload any image.

I tend to disagree here... Based on the documentation from Apple, they are waiting to get *several* matches, *not only one* (we don't know what is *several* but I don't expect something like If yes, after the manual process, the authorities are called.

Hypothetically, what happens if a viral event should persuade people to mass upload these images? Would Apple ad hoc modify their review protocol?
Post reply on HN