this post was submitted on 17 Jul 2026
197 points (98.5% liked)
privacy
10404 readers
6 users here now
Big tech and governments are monitoring and recording your eating activities. c/Privacy provides tips and tricks to protect your privacy against global surveillance.
Partners:
- community.nicfab.it/c/privacy
founded 4 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
But does it work?
Still not sure.
Untested means, unlikely to work at all.
Maybe the IR LEDs at night.
Gets run over by a tesla...
:-D
My first thought when I saw this - "oh hey look, razzle dazzle clothes" which is my personal mental association of the famous scene in Stripes movie, with Dazzle Camouflage which I always thought was a neat idea.
I’d imagine it’s a bit of cat and mouse though, because if they train the drones to target that pattern they’d be sitting ducks
From my understanding (which could be incorrect, I mean I wasn't there), the pattern applied to the hull was done differently per ship. I imagine they applied the base pattern in a random tile-like build where things can be rotated and moved around in the design before painting.
just from my reading of papers, optimized patterns make use of the fact that detectors (even modern transformer based architectures) are capital S Sensitive to the sillouette of an object.
in my other comment i linked a paper where they fool modern detectors with 80+% success rate. the clothing looks like a tiedye shirt to me. not fashionable, but not drawing attention either. if there is a way that you can make this fashionable is not a question i can answer :p
from my reading combining detectors (ensembling) does not help. this leaves spies with a problem, how to detect these noisy people without losing performance on normal nonnoisy people. there are tricks to this, but they are limited at best, for the same reason the noise worked in the first place. here is the reason.
assume noise makers create noise with methods A,B,C. you train on a dataset with images from the different noise makers (you don't know which method they used). each of these noisy groups will have a distict sillouette you can detect, but together, you are building a function like:
pineapple: fruit bannana: fruit apple: fruit
this is a solvable problem for ml, so how do the noise makers win ultimately? randomization in the process. if each noise is distinct there is no training, you only have one image per type of fruit, you might as well use traditional CV to detect these noisy people (good luck).
My opinion is i want a product where the customer provides a random seed which generates the process of noise generation for their shirt. the product is the transparency into how the seed effected the noise making, and an evaluation of the noise on modern off the shelf detectors.
obviously we don't see that here, but it seems doable, maybe a business idea for me? haha
Thanks for the thoughtful reply. I like this idea esp.:
Seems pretty doable in the age of printed t-shirt websites etc.
Or just, you know, more diversity in clothing. I bet surveillance ml gets a massive advantage from people's tendencies to stick to some sort of norm - aka fashion.
In any case none of this actively opposes the huge datacenters sucking up land, water and electricity first in the US, soon all over the planet ... and that's what saddens me. I don't know what to do about it either though.
Thing is, I feel like having bright dazzling lights on your hoodie just fingerprints you even more, no?
Maybe if you use it strategically it could help, but I wouldn’t be sold on that feature
I don't think that's a problem, but I suppose it depends on your goals. Facial recognition in low-light conditions is already hard enough of a problem that they, presumably, are not building IR-light-clothing detection into their algorithms. And if IR-blasting clothing becomes commonplace enough that they start working on it, then it's also no longer distinctive. So, if the goal is to defeat passive surveillance of your movements, this should be highly effective.
If a human reviews the footage, then none of this likely matters anyway, since they'll be able to identify other identifying features about you—but that's a scalability problem. You'd need to have done something very "interesting" to be worth investing the time into tracking across hundreds of video feeds through a surveillance network.
If your goal is to get away with a crime or actively resist detection while protesting, then this is (obviously) a bad idea to wear this. You should instead choose something as non-descript as possible, like all black, long sleeves, gloves, face mask, and tinted goggles. But you can't just walk around in public like that, in general.
Translation is they don't work.
They don't have a lot of technical details but essentially what these do is try to look maximally like 'eyes' or whatever other targets are used by facial recognition to some AI model.
They have to choose which model(s) to train against and there is no guarantee that the same output would fool a differently trained model.
The IR LED dazzlers should work against anything that is using IR as they work by overloading the sensor. The issue is that most cameras are not using IR in most conditions and will switch to visible wavelengths with enough light.
You could technically do the same thing with visible lights, but you're going to annoy people if you're walking around in a hoodie ringed with 1,000 lumen LEDs.