Not to mention how differently it reads if it starts with: "First they came for the child molesters..."
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 Provides Further Clarity on Why It Abandoned Plan to Detect CSAM in iCloud Photos
- Thread starter MacRumors
- Start date
- Sort by reaction score
Not to mention how differently it reads if it starts with: "First they came for the child molesters..."
If they use a source code compiler they're no different than microsoft, at that point you might as well get Windows.
Cool, go read the paragraph yourself - it’s not saying what you claim it’s saying.
Also, you only use quotation marks when you are actually quoting something.
Again if there is ZERO concerns, why is there even a threshold on false positives at all. If this is truly how it functions there should be absolutely zero false positives. So there is no need for a threshold.
“Visually identical” yes it is saying what I am. Things can be visually identical but not the same.
Check my original post before you do that. I was quoting it. This is what I said.
Straight up from the white paper. I can’t copy it verbatim from my phone for some reason. But check the NeuralHash section.
Last edited by a moderator:
But we don't know to what degree a manipulation would be "safe". And the fact that there was a workflow, threshold and potential false positives BY APPLE, says it's not a 100% perfect match. The presence of false positives and reaching a threshold MEANS there will be some judgment. Otherwise it would be absolutely impossible.....no questions....as to a match.
I did. If it works as you all state, that there is literally ZERO possible chance ANY image would produce the same hash, then there would be NO false positives thus no need for a threshold mechanism. But the pretense of such system means there WILL be SOME false positives.....what are false positives? Oh images that are not exact! Therefore it makes some judgements. The presence of how Apple was handling false positives is proof of what I am stating. Because if it DID NOT do that, there would be 0.000000000000000% chance of ANY false positives. Not even 0.000000001% chance of false positives.
You clearly haven’t looked nearly hard enough
Not a single quote that you have provided is a quote from the document, you have paraphrased at best and in the last quote in a very misleading way since it says “identical and visually similar”.
Give ‘em an inch…
New York police will use drones to monitor backyard parties this weekend, spurring privacy concerns
The New York City police department plans to pilot the unmanned aircrafts in response to complaints about large gatherings, including private events, over Labor Day weekend.
www.nbcnews.com
Yes it is! I said the words "visually similar images"
EXACT QUOTE.
The main purpose of the hash is to ensure that identical and visually similar images result in the same hash, and images that are different from one another result in different hashes. For example, an image that has been slightly cropped or resized should be considered identical to its original and have the same hash.
I was on my phone at the time and could not copy and paste this. So I just switched tabs for a section of words to make sure I said it exactly. Now I am on my Mac, I was able to copy/paste the above. Look....."visually similar images" is in there.
This was my original post on it.
I missed the second word "main" and I changed "the hash" to "NeuralHash" to make it clear in the post. But it essentially the same as the white paper. I even stated a disclaimer that I couldn't copy and paste it so give me a bit of a break....geez.
Things can be visually identical but not the same. What you're saying is that someone might take a photo that is visually identical to some of the worst know sexual abuse imagery of children, but it's not the same photo, it's a new one.
Is that really the argument you want to make here?
The point is that the false positive won't generally look anything like the target CSAM image. These two images have the same NeuralHash:
So do these:
In both cases images of dogs were used as the target database and other images were manipulated until they matched the hash-- so the false positive isn't a natural image it was modified until it matches (notice the color splotches on hte lower car image).
If you notice, a false positive doesn't mean you have another image similar to a dog and in the CSAM case it wouldn't mean you have a picture anything like a CSAM image.
So the reason to set a false positive threshold is to account for the fact that hashes are data reductive by definition and you can sometimes hit a match by accident. If that were to happen 30 times in images you were uploading to iCloud, then it would trigger a manual review- not of your images but of an encrypted derivative of the image. I'm not sure what that derivative is but presumably it's not the whole image because they don't want their reviewers to be subjected to the image from true positives.
So if it's a car rather than previously known child abuse, the manual reviewer would reject it. If it's a loving parent taking a photo of a baby in a bathtub rather than previously known child abuse, the manual reviewer would reject it. It if was previously unknown child abuse rather than previously known child abuse, the manual reviewer would reject it.
If you're unlucky enough to get flagged for an image that is visually identical to, but is not, previously known child abuse then your lawyer has their work cut out for them.
Last edited:
Yes that is what I am saying. Why wait until 30 flags? If this is so perfect as you are all saying, the FIRST occurrence of this should get flagged to the authorities because it CAN'T make false positives. You all are saying how perfect this system is and our concerns are "ridiculous" but then why not just set the threshold at 1 then? The presence of a threshold system and Apple's own words about a very rare false positive chance proves our concerns. That is the entire definition of a false positive.
And it's probably good that my ignorance is showing because I have NO CLUE what these images look like. My main concern is I know people that are in a consenting relationship and are adults that share images. But some bodies are different. And CSAM probably includes some "mature" looking 16/17 year olds posing and some "younger" looking 22-25 year olds could have matching bodies that would get flagged POSSIBLY.
And thank you for actually answering my question instead of nitpicking on my words when I clearly left a disclaimer. Appreciate it!
Last edited:
You’ve claimed to have spent so much time reviewing this and yet you still make these posts 🤷🏼♂️
You are identifying things that will not be flagged.
I have. The person I replied to EVEN PROVED what I found. A car and a dog matched. And they aren't even visually similar!
Then why even have a manual review process then if it absolutely cannot flag false positives? If you get one hit you should go to jail....right? If you say no, then there is a chance of false positives. Which by definition means content in question doesn't match the database.
No, they said EXACTLY what I am saying. THEIR QUOTE.
What are these examples given by SOMEONE ELSE? False positives. Exactly what I am saying. Why have a manual review process AT ALL then if ITS NOT POSSIBLE? One hit should send you to jail then.
Read what that person said to you again.
Also whilst you are at it, go read some technical papers on Neural Hashing.
Last edited by a moderator:
I did read. You all are saying there is no chance for false positives. But this person said the following.
So if it's a car rather than previously known child abuse, the manual reviewer would reject it. If it's a loving parent taking a photo of a baby in a bathtub rather than previously known child abuse, the manual reviewer would reject it. It if was previously unknown child abuse rather than previously known child abuse, the manual reviewer would reject it.
Last edited by a moderator:
Where have I said there isn’t a chance of a false positive? Please point me to exactly where I said that, I will wait.
“other images were manipulated until they matched the hash-- so the false positive isn't a natural image it was modified until it matches (notice the color splotches on hte lower car image).
If you notice, a false positive doesn't mean you have another image similar to a dog and in the CSAM case it wouldn't mean you have a picture anything like a CSAM image.”
You are fighting me about the entire process of false positives. Saying I am wrong and don't know how it works. Yes you are battling me about false positives.
Regarding the image.....because it was manipulated to show the example that a car and a dog can match. They would NOT demonstrate a known CSAM image....that would be illegal. So obviously it would not be the same example. And it still shows a car can match with a dog.
Not one statement I have made was incorrect if false positives are even POSSIBLE. The presence of a manual review process proves there are potential for false positives.
But you all just talked down to us saying that is NOT how the system works and claim we didn't do our research when we did. Other posts showed and supported it! My ENTIRE concern was the false positives.