acm-header
Sign In

Communications of the ACM

ACM TechNews

Can a Stealth T-Shirt Foil Digital Spies?


View as: Print Mobile App Share:
Northeastern University professor Xue Lin wears the adversarial shirt.

Researchers at Northeastern University, the Massachusetts Institute of Technology, and IBM have designed stealth apparel that renders the wearer invisible to spying computers.

Credit: Lane Turner

Researchers at Northeastern University, the Massachusetts Institute of Technology, and IBM have designed stealth apparel that renders the wearer invisible to spying computers.

The adversarial T-shirt sports patterns that confuse artificial intelligence (AI) systems so they cannot identify the wearer as a human being.

The shirt exploits AI vision systems' perception of pixels, the addition of which can fool the systems.

Northeastern's Xue Lin and Kaidi Xu engineered the shirt to trick the Yolo object-recognition algorithm, which did not identify the clothing's wearers as people 63% of the time.

Said Lin, “We try to explore the vulnerability of these neural networks, and hopefully, we can fix this problem.”

From The Boston Globe
View Full Article - May Require Free Registration

 

Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA


 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account