“You can often hide from an AI video system with the aid of a simple color printout.”

Hiding in plain sight


Mon, Apr 29th, 2019 09:51 by capnasty NEWS

MIT Technology Review brings to attention this study from the Belgian university KU Leuven that demonstrates how easily video tracking technology can be fooled with a simple coloured printout.

In this paper, we presented a system to generate adversar-ial patches for person detectors that can be printed out andused in the real-world. We did this by optimising an imageto minimise different probabilities related to the appearanceof a person in the output of the detector. In our experimentswe compared different approaches and found that minimis-ing object loss created the most effective patches.

From our real-world test with printed out patches we canalso see that our patches work quite well in hiding personsfrom object detectors, suggesting that security systems us-ing similar detectors might be vulnerable to this kind of at-tack.

We believe that, if we combine this technique with asophisticated clothing simulation, we can design a T-shirtprint that can make a person virtually invisible for automaticsurveillance cameras (using the YOLO detector).



You may also be interested in:

“Many people say they’re resigned to the idea that we’re going to be spied on and there’s nothing we can do about it.”
"One thing sits at the heart of what many consider a surveillance state within the US today."
"Cell phone numbers are better identifiers than email addresses."
Signal: Free, Worldwide, Encrypted Phone Calls for iPhone
How on Facebook Everyone Can Have Your Private Phone Number