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

Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

github.com

Post reply on HN
Print view View original

131–140 of 363 posts

Re: Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

#131

Post by cirrus3 »

Alupis wrote: post #35

So, what does Apple get out of all this, except negative attention, erosion of their image, possible privacy lawsuits, etc? I just don't understand what Apple's motivation would have been here. Surely this fallout could have been anticipated?

Possible avoidance of being told how they have to do it later.

They will have to do it either way, and they the fact they are even telling how us they plan to do it is more than we can say for every other cloud services.

This is better than all alternatives at this point. Like it or not. If you don't like, you might need to get up to speed on what other services you may already be using are doing.

Re: Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

#132

Post by cirrus3 »

throwaway384950 wrote: post #67

Some people seem to be confused why a hash collision of a cat and a dog matters. Here's a potential attack: share (legal) NSFW pictures that are engineered to have a hash collision with CSAM to get someone else in trouble. The pictures are flagged as CSAM, and they also look suspicious to a human reviewer (maybe not enough context in the image to identify the subject's age). To show that this can be done with real NS…

You seem to be assuming a human cannot tell the difference from some random NSFW content, and some legit known CSAM, 30 times. Try again.

Re: Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

#133

Post by heavyset_go »

shuckles wrote: post #127
SCLeo wrote: post #125

Earlier quoted context omitted.

Oh, I think I misunderstood you. I thought you meant instead of "sending images that collides with perceptual hashes of known CASM", why not "send actual CSAM in 'Facebook Messenger or Gmail or Dropbox', and since those services also use some other detection algorithm, it will also incriminate the receiver."

Those services will take your account through the same, if not more invasive, process if you are found with a hash match like the ones being proposed in these comments. Unlike Apple, they’ve built interfaces that surface all your account activity to reviewers.

> Unlike Apple, they’ve built interfaces that surface all your account activity to reviewers.

You can't know this without independent audits.

Re: Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

#134

Post by onepunchedman »

norov wrote: post #114
onepunchedman wrote: post #76

Earlier quoted context omitted.

Nah that's so 2020, 2021 is all about low resolution legitimate porn being transformed to match CSAM. Get with the times!

But why low resolution porn?

So that you are able to bypass the manual reviews. It still looks like CSAM, but it isn't.

Re: Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

#135

Post by cookiengineer »

onepunchedman wrote: post #25

This is just getting wilder and wilder by the day, how spectacularly this move has backfired. As others have commented, at this point all you need is someone willing to sell you the CSAM hashes on the darknet, and this system is transparently broken. Until that day, just send known CSAM to any person you'd like to get in trouble (make sure they have icloud sync enabled), be it your neighbour or a political figure, an…

Imagine being a parent that made pictures of their own children that bathed naked in their own backyard.

I don't know about you, but my parents certainly have lots of embarassing pictures of me in their photo album.

There will be so many false positives in that system, it's ridiculous. It doesn't necessarily have to be a false colliding hash, but legitimate use cases that - by definition - are impossible to train neural nets on unless the data is being used illegally by Apple.

Re: Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

#136

Post by endisneigh »

Dylan16807 wrote: post #92
dragonwriter wrote: post #64

Earlier quoted context omitted.

> Apple's reviewers, by law, cannot look at the target This is false. > No one except NCMEC is allowed to possess the target (CSAM material). False. No one is allowed to knowingly possess it, without taking certain actions forthwith when they become aware of it. Obviously, prior to it being reviewed as it is , neither the reviewer nor Apple has knowledge that it is actual or even particularly likely CSAM.

I think you misread what was meant by "target". Yeah, Apple might be able to look at the uploaded image. But the reviewers don't have a copy of the original image added to the database , which is the "target".

You’re correct, but the uploaded image is sufficient as it would be obvious some that it isn’t CSAM material.

If it was, then would it matter if it wasn’t the original?

Re: Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

#137

Post by simondotau »

copperx wrote: post #75
robertoandred wrote: post #58

Earlier quoted context omitted.

What kind of social engineering would lead an innocent person to save known CSAM to their photo library?

None needed. You could just send a photo to the target through WhatsApp, and the photo would be automatically synced with iCloud.

Wouldn't the photo be scanned for CSAM by WhatsApp first?

Re: Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

#138

Post by onepunchedman »

copperx wrote: post #94
onepunchedman wrote: post #89

Earlier quoted context omitted.

Yeh I was talking in hyperbole, but the possible attack vectors this system enables are so powerful I felt it warranted. Under this system you are able to artificially ddos organizations that verify if CP is sent by sending legitimate, low-res porn whose hash has been modified. You can trigger legitimate investigations by sending CSAM through WhatsApp or through social engineering. You can also fuck with Apple by sen…

And this is just one step away from Apple and Microsoft building this scanning into the OS itself (into the kernel/filesystem code, why not?!). This is beyond insane. Stallman was right. Our devices aren't ours anymore. Now, to be fair, there would be a secondary private hash algorithm running on Apple's servers to minimize the impact of hash collisions, but what's important is that once a file matches a hash locally…

Jup. As others have pointed out, if Apple were willing to lie about the extent of this system and its inception date, why should we suddenly trust that they won't extend its functionality. They themselves explicitly state that the program will be extended, so if this is the starting point I don't think I will be around for the ride.

It's a shame as I really love some of their privacy-minded features (e.g. precision of access to the phone's sensors and/or media).

Re: Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

#139

Post by XorNot »

ComputerGuru wrote: post #102
XorNot wrote: post #54

Earlier quoted context omitted.

The "visual derivative" is not something any of us have been shown an example of either. Whatever it is, I suspect you only need to be vaguely in the same ballpark (I would wager humanoid shaped skin tones maybe). So I suspect it would be easier then that (particularly since this whole hashing scheme has been surrounded with a lot of clear garbage - "1 in a trillion" -> on demand collisions in a couple of weeks?

I think visual derivative is just a beating-around-the-bush way of saying “thumbnail.”

That's their words that they feel no need to elaborate on. Obviously they actually seem to just be doing the "technically the truth" thing - which shows that someone realized no one would like hearing what it actually is.

Re: Show HN: Neural-hash-collider – Find target hash collisions for NeuralHash

#140

Post by ec109685 »

bArray wrote: post #48
Scaevolus wrote: post #17

NeuralHash collisions are interesting, but the way Apple is implementing their scanner it's impossible to extract the banned hashes directly from the local database. There are other ways to guess what the hashes are, but I can't think of legal ones. > Matching-Database Setup. The system begins by setting up the matching database using the known CSAM image hashes provided by NCMEC and other child-safety organizations.…

To defeat this, all you need to be is a state actor with a database of child porn at your disposal (which is stored for exactly the purpose of training detection systems). Then you run the hashing algorithm against images you know are in the database (Apple suggested that they would accept suggestions by some kind of multi-Country vote). Then you can pull out the hashes and figure out how to trigger false positives o…

This doesn’t work for two reasons: 1) There’s no way to know the perceptual hash value of Apple’s private NeuralHash function that is run on the derivative of the image server side to verify a hit really is CSAM. So while you could cause a collision with the on device neural hash if you possessed illegal content, you wouldn’t know if you successfully faked Apple’s private neuralhash implementation. 2) An Apple reviewer must verify the image is illegal before it’s passed along to law enforcement.
Post reply on HN