Live data from Hacker News

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

github.com

1–10 of 363 posts

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

#2
The README (https://github.com/anishathalye/neural-hash-collider#how-it-...) explains in a bit more detail how these adversarial attacks work. This code pretty much implements a standard adversarial attack against NeuralHash. One slightly interesting part was replacing the thresholding with a differentiable approximation. I figured I'd share this here in case anyone is interested in seeing what the code to generate adversarial examples looks like; I don't think anyone in the big thread on the topic (https://github.com/AsuharietYgvar/AppleNeuralHash2ONNX/issue...) has shared attack code for NeuralHash in particular yet.

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

#4
post #3

That was scary fast. Is there a point in using this algorithm for its intended purpose now?

If the intended purpose is to lead by example and eventually mandate code on every computing device (phone and computer) that scans all files against a government provided database, then yes, that purpose still exists and this algorithm still works for it.

Just wait and watch - I guarantee you that Apple will be talking about CSAM in at least one anti-trust legal battle about why they shouldn't be broken up. Because a walled garden means they can oppress citizens on behalf of governments better.

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

#7
Ok so now all we have to do is get a phone, load it with adversarial images that have hashes from the CSAM database and we wait and see what happens. Basically a honeypot. Get some top civil rights attorneys involved. Take the case to the Supreme Court. Get precedence set right.

Lawfare

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

#8
post #6

Hi, this looks interesting but I have no idea what this all means lol Is this some way of hiding a picture within a picture, or am I way off the mark?

Apple began scanning for CSAM with Neuralhash. This allows you to turn an image into a specific neuralhash thus possibly triggering its (automatic) CSAM detection. Imagine if a picture of a cat could cause Apple to think you have CP on your device.

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

#9
post #6

Hi, this looks interesting but I have no idea what this all means lol Is this some way of hiding a picture within a picture, or am I way off the mark?

Apple began scanning for CSAM with Neuralhash. This allows you to turn an image into a specific neuralhash thus possibly triggering its (automatic) CSAM detection. Imagine if a picture of a cat could cause Apple to think you have CP on your device.

Oh damn, that’s crazy. Very cool project. Thanks for sharing and the explanation.
Post reply on HN