That’s super obvious how that would be possible.
You’re looking at this with human eyes, not with AI eyes.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Addresses CSAM Detection Concerns, Will Consider Expanding System on Per-Country Basis
- Thread starter MacRumors
- Start date
- Sort by reaction score
You’re looking at this with human eyes, not with AI eyes.
That in and of itself doesn't mean very much. Usually when a security audit is done, it is explained in a report what was tested, under what conditions, and if there were any weaknesses or security holes found etc.
Sorry, nothing contradictory about that.
There’s an information gap with regards to AI, hashes, etc.
An image is just a series of pixels. And the fact that there is some flexibility from the truly stored image in the database and what can get flagged (any modifications get flagged), means there is a wide range of collision. Its not 1:1.
A similar image with nearly the EXACT SAME pixel attributes, but with an adult/legal subject CANNOT get flagged....how? The pixels are nearly the same....
Sure.
They probably won’t let people actually audit it.
Someone brought it up in this FAQ:
FAQ about Apple's Expanded Protections for Children - TidBITS
Apple is piercing the privacy veil on our devices to protect children. The company claims its efforts won’t open up a Pandora’s Box in the interests of averting sexual exploitation of children or recognition of sexual material handled by children under 18 when a parent wants oversight. But it’s...
tidbits.com
The unlucky 1 in a million image that tricks the system by sheer pixel attributes would be something completely unrelated to children, like a picture of a bucket of sand or a sunset. Human review would can it in a split second. Fact is, it wouldn’t even reach human review ‘cause one match isn’t enough, an account has to do multiple offences.
On the other hand, altered versions of the original known CSAM would be easily caught by the AI. (again, read about how PhotoDNA works, Apple’s system is probably similar)
There is no contradiction.
Giggles, time and again you've tried to explain how this works (and in some detail).
Lots of people on here are not listening or reading correctly. Apple's papers on this go into extreme technical detail for those interested and have the appropriate attention span.
I have read it. It makes no sense. Let me put it another way. Using @hans1972 example.
So we are saying in this example, picture 2 will NOT get flagged, even though it is nearly identical.....correct?
Okay, so I fire up Photoshop, and I perform a warp on Picture 1 to make it COMPLETELY MATCH picture 2. However, all these reports are now saying Picture 1 modified (which MATCHES picture 2) will now get flagged (because it tracks crops, edits, manipulations and other things)? But somehow picture 2 does not?
Again, that is the contradiction. ANY modifications to picture 1 get flagged, but picture 2 does not? How?
It is an AI. You do understand that Apple calls their AI component on their processors Apple Neural Engine (machine learning focused processor) and this exactly how it works. They are not looking for exact matches. Looking for exact matches would be idiotic and counter productive. What they do is train the AI to recognise images based on a set of aprox. 200 000 images. If the scanned image falls within threshold then pictures will be flagged and manually checked.
And the funny thing is, the same CSAM technical document that they keep telling me to read (which I did) says this exactly.
Training a neural network is AI.
iMessage photos share to your photos now. Therefore if iCloud is on and someone sends you CP it gets compared to hashes and then you get reported to authorities.
After thinking about this for a couple of days now. I really don't care if they want to scan my photos when they're in iCloud as mentioned several times other companies are already doing this. However, I'm not interested in them scanning photos on my device. And yes, I get that they will only be scanning if I select the option to backup my photos to iCloud but the scan will still be done on my device. Once they implement this technology, it could then be used for other things too; to what extent I don't know. Will hackers find a vulnerability in this and then use it to scan for other things? It's opening another possible point of failure that doesn't need to be done.
Because the technology used is irrelevant. They access and search private images. That‘s called snooping.
Its absolutely irrelevant HOW they do it. Important thing is: they do it
I trust this as much as I trust Ebay
www.nytimes.com
Ex-eBay Workers Sent Critics Live Roaches and a Mask of a Bloody Pig Face, U.S. Says (Published 2020)
Six former employees of the site sent threatening messages and deliveries to a couple after the e-commerce newsletter they published wrote about a lawsuit involving eBay.
My guess the impact to battery life would be negligible; especially on newer devices.
If you really want to see results, vote with your wallet. Even a 10-20% decline in revenue will result in Wall Street instantly taking notice with a declining share price. That will get Apple's attention ASAP.
Yada yada-ing on internet forums will have no effect.
Will you do it?
I already have prior to this. No Apple services subscriptions. No Euros going from me to Apple. I am Android and staying that way. Not recommending Apple products and services to friends or family.
Even thinking this is not a reason to propagate or endorse misconceptions about the technical aspect of this.
Further discussion about the legal/ethical/philosophical/etc. aspects must stem from a common base of truth about the technical reality of what we’re talking about.
My action alone I agree. Unfortunately there are not enough intelligent consumers out there to make a big difference. But I sleep better knowing I am not an Apple customer. As I said earlier, protesting, publicising, and informing people about the way Apple is going is a key part of the equation. Apple will fall, all smug companies do eventually.
The AI in the hash comparison is for catching slightly altered versions of the picture (you proved too obtuse to understand how but that’s another matter).
The AI in the iMessage Safety for kids under 12 is for looking at the content of the photo.
They’re both called “AI” but they do completely different things.
But you’re too limited to acknowledge this too.
And the base truth is: they access and snoop around my data on my device, isn‘t it?
No they won't or they may fall when my grandkids have grandkids.
And is the case seemingly for Apple, they will lead the charge here.
(Not that I am onboard with this, but my opinion to your opinion above)