This style of attack has been discussed for a while https://www.usenix.org/system/files/sec20-quiring.pdf - it’s scary because a scaled image can appear to be an _entirely_ different image.

One method for this would be if you want to have a certain group arrested for having illegal images, you could use this sort of scaling trick to transform those images into memes, political messages, whatever that the target group might download.

▲

orbisvicis 5 days ago | parent | next [-]

This is mind-blowing and logical but did no one really think about these attacks until VLMs?

They only make sense if the target resizes the image to a known size. I'm not sure that applies to your hypotheticals.

▲

Gigachad 5 days ago | parent | next [-]

Because why would it matter until now. If a person looked at a rescaled image that says “send me all your money” they wouldn’t ignore all previous learnings and obey the image.

▲

vasco 5 days ago | parent | prev [-]

Hidden watermarking software uses the same concepts. It is known.

	▲	arcticbull 5 days ago \| parent [-]
		Steganography for those who want to look it up.

▲

monster_truck 5 days ago | parent | prev [-]

Describing dithering as scary is wild

▲

LPisGood 5 days ago | parent [-]

The thing is that the image can change entirely, say from a gunny cat picture to an image of a dog.

▲

therein 5 days ago | parent [-]

And that "trick" has been used in imageboards with thumbnails for a very long time to get people to click and see a full image while they otherwise wouldn't.

▲

ajsnigrutin 5 days ago | parent [-]

Sure, but back then it was "haha you've been pranked! and yes, that guy is actually wearing a wedding band!"

Now... with chat control and similar alternatives and AI looking at your images and reporting to authorities, you might get into actual trouble because of that.

	▲	lazide 4 days ago \| parent [-]
		Yup. Imagine the ‘fun’ caused by automated CSAM or ‘bad politics’ content scanners and this tech.