$400 sweater tricks facial-recognition cams into thinking you're a dog

Patterned after animals, the clothes can protect the user's privacy by causing the camera to mistake them for their fuzzy animal counterpart.
Kavita Verma
The Manifesto Collection.
The Manifesto Collection.


With facial recognition technology becoming increasingly widespread, there's been a steady rise in justified concern about how our data could be abused. 

To protect their privacy and combat this threat, savvy designers create fashion items that can outsmart the tech. CAP_able is an Italian startup that is the latest to join this trend. They are offering fashionable solutions to help safeguard individuals' identities.

A new apparel collection by Italian startup CAP_able is designed to throw facial recognition algorithms for a loop. The debut Manifesto line features hoodies, tops, and dresses boasting an "adversarial patch," which looks like festive Christmas sweaters from afar. But in reality, they are patterned after various animals, including giraffes, zebras, and dogs. 

This clever ploy can either cause the camera system to fail to identify its wearer or mistake them as their fuzzy animal counterpart.

The mindset between engineering and fashion

The idea of making something that can trick facial recognition software struck Didero when she was on a Master's exchange at the Fashion Institute of Technology in New York. She read about how Brooklyn tenants are fighting with their owner's plans to install such a system into their building. 

Coming up with the idea was simple, but to bring it into reality, she had to find and design the "adversarial algorithms" that are capable enough to fool facial recognition software. She said," you must have a mindset between engineering and fashion."

She and her friend, a computer science engineer, had two options to make this dream a reality. Either they could come up with an image of our giraffe and use an algorithm to adjust it. Or they set the size, color, and form they wanted the image to take and let the algorithm do the rest of the work. 

After considerable experimentation, Didero and their team created a process for generating physical garments with images woven onto them. 

Using the popular object detection system YOLO as well as Computerized Knitwear Machines from Italy - they mastered optimal image placing on Egyptian cotton materials to create truly unique clothing items that fooled algorithms like YOLO about 60 to 90 percent of the time. 

Brent Mittelstadt, director of research and associate professor at the Oxford Internet Institute, added with prices starting at $300, anti-surveillance clothing could be the ultimate way to protect your privacy. Unless, of course, it becomes obsolete. 

As Woodrow Hartzog from Boston University School of Law noted, these clothes may do more than just shield wearers against snooping eyes; they can also reduce stigma and help encourage lawmakers to create meaningful rules. 

While this type of apparel might only appeal to a select few right now, its potential benefits suggest that it is an innovation well worth watching in the future!

Future Improvements 

CAP_able launched a Kickstarter campaign late last year, and they raised about €5,000. In order to refine their business model, they are now planning to join Politecnico's accelerator program