Hacker News Reader

Hacker News

Comment by marcus_holmes | Hacker News Reader

Full thread

marcus_holmes

I don't understand how this is supposed to do anything except generate false positives.

If someone takes a new picture of a child being abused, that won't be on the database so won't get flagged.

Are they expecting child-abusers to take photos of existing photos of children being abused?

Or are they hoping to catch abusers sending each other known photos of child abuse unencrypted? Because that's:

a: stupid (the even-remotely-smart abusers will encrypt their abuse) and

b: rife for mis-use (I, using a non-Apple device, send your Apple device a photo of abuse. You're now flagged as an abuser and your life shuts down)

As others have said, I don't think this has anything to do with child abuse, because it plain doesn't work as a method of preventing child abuse. It totally works for flagging "politically unacceptable" images, though. Maybe the CCP got fed up of all those memes?

View on HN

I'm surprised how shocked people are about this new thing from Apple. Both Google and Microsoft are already doing this stuff and have been since 2014[0][1]. How come Apple is getting such a huge amount of hate for this, but nobody even seems aware of the others already having implemented such technologies.

I also hope nobody here ever gets mad about the rampant child pornography going around on platforms like Kik, because that would be incredibly hypocritical.

0: https://www.theverge.com/2014/8/5/5970141/how-google-scans-y...

1: https://nakedsecurity.sophos.com/2014/08/10/microsoft-scans-...

It has to do with the way Apple advertises being a herald of privacy.

Also your links say that Google and Microsoft scan e-mails on the cloud. Not phone storage as in Apple's case.

> Also your links say that Google and Microsoft scan e-mails on the cloud. Not phone storage as in Apple's case.

Another confirmation that Google and MS have access to the data you store in the cloud and Apple does not. It's very unfortunate that scanning this locally is viewed as a negative in stead of a positive.

What confirmation? Apple already has access to your cloud data:

FBI 'persuaded Apple to halt iCloud encryption' - https://www.bbc.com/news/technology-51207744

Yeah, you expect it from Google and MS; Apple was the one tech giant that promised to value your privacy, so now that they've turned on that, there are no mainstream options left. Hence the strong dismay.

(Honestly Apple probably doesn't really have an option -- the govt wants this, and I'm sure they can exert all kinds of leverage to get it if Apple doesn't comply.)

Unless Apple is starting to view YOUR phone as THEIR property - this is an invasion.

Android does not scan your phone. Google scans content uploaded to ITS machines - because content hosted on ITS machines is a liability.

Right and how you reduce that liability is to do the scanning client-side and reject the upload/send. Google does this with Drive uploads and WhatsApp does it on message sends.

You can't be in hot water for hosting/distributing CSAM if it never gets on your servers/network in the first place.

I was talking with my roommate about this yesterday. Similarly to how in the article there was a cat picture that the NN said was guac, I wonder if we will start to see the rise of "booby trapped images" where you might be downloading a wallpaper but unknown to you, that image contains CSAM hashes.

Hash collisions are theoretically impossible. It's a completely different issue than a NN misclassifying an image. It reflects poorly on the author that they made this argument.

I'm sorry, but this is 100% incorrect. In fact the opposite is true.

Hash collisions are 100% impossible to avoid. You are mapping an infinite set (the set of all possible images) to a finite set (a fixed length number).

Cryptographic hashes are designed so that collisions are hard to construct at will. But this is not a cryptographic hash at all and I wouldn't be surprised that constructing an image that matches a given hash is easy.

>I'm sorry, but this is 100% incorrect. In fact the opposite is true.

>Hash collisions are 100% impossible to avoid. You are mapping an infinite set (the set of all possible images) to a finite set (a fixed length number).

If you want to be needlessly pedantic I guess. But for 99.999999% of usage hash collisions don't exist in practice.

>Cryptographic hashes are designed so that collisions are hard to construct at will. But this is not a cryptographic hash at all and I wouldn't be surprised that constructing an image that matches a given hash is easy.

You got a cite to back this up? Because they claim otherwise.

> If you want to be needlessly pedantic I guess. But for 99.999999% of usage hash collisions don't exist in practice.

Oh right, that very soothing. The very device I spent $1000 on has 0.0001% of ruining my life by causing a no-knock raid due to a false positive. They should put this stuff on their ads, makes me wanna buy more Apple products.

I'd much rather sell all my current Apple devices and permanently switch to linux than do this.

No, it's one of the 99.999999% of times where there's 0 risk.

I'm pointing out that it is technologically possible to do so, and I do not want to bet my life that Apple's code has no bugs. I've been writing code for 20 years, no one can convince me this shit won't ruin at least one life in the next decade after its release.

There are probably a billion Apple devices out there [0]. Each of those devices has at least a couple hundred unique images on it [1].

If you're correct, and the chance of a hash collision is 0.0001% then for each hash in the database, that's 200,000 collisions.

Assuming the database has 1,000 hashes [2] and there's no overlap in collisions for any given person, that 200 million peoples' lives ruined.

[0] rough guess based on: https://www.statista.com/statistics/276306/global-apple-ipho... the exact number might be off but I think the order of magnitude is about right.

[1] My gf has taken at least 3 photos every day for the last 5 years at least (so well over 5000 unique photos) - 200 is very conservative.

[2] Again, very conservative. This is a collection of all known child porn images. I wouldn't be surprised if the number is actually two orders of magnitude higher.

[edit] replied to the wrong comment. Bugger. Hopefully contributed to the conversation anyway.

Apple says:

>less than a one in one trillion chance per year of incorrectly flagging a given account

If we gave every single human being an iPhone we'd expect an incorrect flag every 160 years or so.

In all fairness, we've all seen marketing twist developer words many, many times. "less than one in one trillion" sounds like a very neat, nice number that somebody decided to fixate on.

This.

And I guess we'll find out.

The horrible thing is, of course, that this whole process will probably be automated and there'll be no recourse to a human with any power to work things out properly. If there are thousands of cases, they'll have to take to Twitter to shine some light on it. Except that in this case they're self-identlfying as suspected child abusers. How many people are going to risk the negatives that go with that? Will we ever find out how many people were actually false positives?

If you are talking about cryptographic hashes, then yes. But perceptual hashes are design to be similar if the images are similar. I still have to read the tech paper, but most likely it uses some sort of perceptual hashing or AI recognition.

>If someone takes a new picture of a child being abused, that won't be on the database so won't get flagged.

Hah. That's an interesting point.

If anything you are encouraging new child porn.

It's a measure to attack the distribution and consumption of child porn, which is a driver for its production. It can also help provide evidence in unearthing child porn rings. Seems obvious.

https://en.m.wikipedia.org/wiki/Four_Horsemen_of_the_Infocal...

There's way more to it than appears at first blush. This is much more than a kiddie porn thing, it's a fundamental shift in the Overton window.

Reply on news.ycombinator.com