Re: Apple accused of underreporting suspected CSAM on its platforms

Jolly Roger <jollyroger@pobox.com> wrote:

On 2024-07-29, Chris <ithinkiam@gmail.com> wrote:

Jolly Roger <jollyroger@pobox.com> wrote:

On 2024-07-28, Chris <ithinkiam@gmail.com> wrote:
 

No-one is going to be charged for a dubious photo of their own
child. There are much bigger fish to fry and get into jail.

 
You're wrong. It has already happened:
 
A Dad Took Photos of His Naked Toddler for the Doctor. Google Flagged
Him as a Criminal <https://archive.is/78Pla#selection-563.0-1075.217>

 
I explicitly said "charged". No-one got charged. The law is working
just fine. It's the tech, as I've been arguing all along, that's the
problem.

 
So it's okay that these parents and their child had their privacy
violated, their child's naked photos added to the CSAM database, and
their accounts (along with years of emails, photos, and so on)
revoked/deleted because they were officially charged?

Nope. Not ok. I didn't say otherwise.

I'm explicitly saying the tech has to be used proportionately and
effectively. The current systems are neither.

Read the whole article to get a glimpse of what innocent people go
through who fall victim to this invasive scanning.
 
Do you think these parents and their child consider their privacy to
be violated? How would you feel if your intimate photos were added to
the PhotoDNA CSAM database because they were incorrectly flagged?

 
This wasn't PhotoDNA, which is what Apple was similar to. It was
google's AI method that is designed to "recognize never-before-seen
exploitative images of children" which is where the real danger sits.
 
It is designed to identify new abuse images based on only the pixel
data so all hits will be massively enriched for things that look like
abuse. A human won't have the ability to accurately identify the
(likely innocent) motivation for taking photo and "to be safe" will
pass it onto someone else make the decision i.e. law enforcement. The
LE will have access to much more information and see it's an obvious
mistake as seen in your article.

 
Actually, a human being does review it with Google's system:

I was unclear. I'm not saying a human doesn't review, I'm saying that given
the dozens/hundreds of images of suspected abuse images they review a day
they won't have the ability to make informed decisions.

---
A human content moderator for Google would have reviewed the photos
after they were flagged by the artificial intelligence to confirm they
met the federal definition of child sexual abuse material. When Google
makes such a discovery, it locks the user’s account, searches for other
exploitative material and, as required by federal law, makes a report to
the CyberTipline at the National Center for Missing and Exploited
Children. 
---
 

Apple's system was more like hashing the image data and comparing
hashes where false positives are due to  algorithmic randomness. The
pixel data when viewed by a human won't be anything like CSAM and an
easy decision made.
 
What's crucial here is that Google are looking for new stuff - which
is always problematic - whereas Apple's was not. The search space when
looking for existing images is much tinier and the impact of false
positives much, much smaller.

 
Yes, but even in Apple's case, there's a small change of a false
positive patch.  And were that to happen, there is a danger of an
innocent person's privacy being violated.

In every case there's a chance of FPs. Apple would have had lower FPR then
*the current* system.

Given the choice I'm in favour of the better, evidence-based method. You're
in favour of the worse system that is actually being used and then complain
that it's worse.

How many children are you prepared to be abused to protect YOUR
privacy?

 
Now you're being absurd. My right to privacy doesn't cause any
children to be abused.

 
That's what you'd like to think, yet the reality is that not only is
it harder to identify perpetrators, but also, ironically, your
position ensures more people get erroneously labelled.

 
Nope. I don't support any scanning of private content.

Yet it's already happening so why not support the better method?

Apple was wise to shelve this proposal. And I am happy to see that
they embraced more private features such as the Safe Communication
feature which is done without violating customers' privacy.

 
It wasn't violating anyone's privacy. For the umpteenth time. It
actually preserved people's privacy by design.

 
While it went further than the rest to protect people's privacy,
there was still room for error and for innocent people to be harmed.
That's the truth you seem to want to disregard, and that's why Apple
shelved it.

 
My truth is that the Apple method was a significant improvement.i

 
I agree - still not good enough for me though.

"Perfect is the enemy of the good"

By seeking perfection you and others are allowing and enabling child abuse.

Plus if people didn't like it - unreasonably - they didn't have to use
icloud.

 
True.
 

Apple only shelved it for PR reasons, which is a real shame.

 
You don't know all of Apple's motivations. What we know is Apple shelved
it after gathering feedback from industry experts. And many of those
experts were of the opinion that even with Apple's precautions, the risk
of violating people's privacy was too great.

That wasn't the consensus. The noisy tin-foil brigade drowned out any
possible discussion. Apple should have simply implemented it like google
are doing, but badly.

What you're disregarding is that the alternative hasn't been more
privacy for people, it's been less privacy and more errors. A
lose-lose situation.
 
You want to live in a world that doesn't exist.

 
Nah. I simply choose not to engage with those alternatives for things I
consider private.

Head-in-sand mentality.