If you’ve ever watched a runway fashion show, you could be forgiven for feeling confused. This is the fashion world satirized by Zoolander and Sacha Baron Cohen’s character Brüno; where wearability and even sartorial attractiveness is deprioritized in favor of crazy designs and artistic boundary pushing. Kate Rose’s clothing line Adversarial Fashion is as confusing as anything you’ll find on any catwalk in the world. The only difference: it’s not you that it’s designed to confuse. It’s the machines.
Adversarial Fashion’s big hook is creating apparel that’s adorned with designs intended to disrupt surveillance systems. If you’ve ever strained to read a tagline on someone’s t-shirt, only to realize it was kinda stupid and you were wasting your time, this same idea is used by Adversarial. Except instead of wasting your time, it’s designed to waste the time and computational firepower of automated license plate readers.
The creative force behind Adversarial (its name is based on the A.I.-befuddling term “adversarial examples“) is a computer security professional and amateur fabric maker named Kate Rose. Not too long ago, Rose was having a conversation with Dave Maass, senior investigative researcher at the Electronic Frontier Foundation. They were discussing the rise of automatic number-plate recognition technology, which has been widely adopted in the U.S. at city, county, state and federal levels.
The American psychologist Abraham Maslow once pointed out that to a person who has only a hammer, there’s a tendency for everything to look like a nail.
This spy tech is meant to track people’s whereabouts by not only recognizing license plates, but also linking this with specific GPS locations. As a person drives around and is captured by multiple cameras, automatic number-plate recognition systems provide a constantly-updated means of monitoring their approximate (or even specific) location. Only according to Maass, it doesn’t work all that well.
The spark of an idea was born.
Creating her designs
“He happened to mention that when they were analyzing data from plate readers, it became clear that the data was not very accurate,” Rose told Digital Trends. “It was pulling in data from misclassified things like billboards and picket fences. There was just a whole lot of junk in there. I thought that, if the specificity on the technology was really that low, maybe I could find a way to create a fabric that could purposely cause similar confusion.”
Rose went away and began carrying out some experiments. “I managed to figure out what the least amount of information that the camera can take in that would still make it believe it was looking at a license plate,” she said. Bringing together her interest in fashion and computer security systems, she began toying around; creating designs intended to befuddle the most diligent of machines.
The American psychologist Abraham Maslow once pointed out that to a person who has only a hammer, there’s a tendency for everything to look like a nail. Well, to an A.I. that’s trained to view the world in terms only of potential license plates, it turns out that they don’t need all that much coaxing to make a mistake.
“It’s kinda fun to play around with it,” Rose continued. “You can see what looks cool from a fashion perspective, in terms of something that someone would want to wear.” Once her designs were completed, Rose set to work making them first into fabrics, and then those fabrics into clothes. To date, she has created a number of Adversarial Fashion items, ranging from shirts and hoodies to jackets, skirts and dresses. These are available for purchase on her website, ranging in price from around $25 to $55, depending on the item.
Broadening the conversation
But is this actually causing disruption or is this more of a political statement in the protest fashion vein of, say, a “F*** Donald Trump” t-shirt?
“Part of this is just about teaching people how the technology works and improve their understanding about part of the surveillance state,” she said. “But it does really work. Even if it’s just one person wearing one of these, and it causes these cameras to ingest 10 to 50 plates which are total garbage, that’s doing something. This is still information that has to be stored and analyzed, all of which costs money. It makes the system slightly less useful. I would love to think of ways that this could be adopted by whole communities who say that they don’t want to be compliant with these systems.”
Rose thinks that clothing items such as this can play a key role in the fight back against mass surveillance.
Ultimately, Rose views what she is doing as helping to prompt a conversation. As she points out, the use of mass surveillance technologies are still widely unknown to many people. Their invisibility is a feature that, well, makes them unnoticed by many. On the surface, at least.
“For example, where I live in Southern California there are malls that you can park in, which tells you where your car is parked when you put your ticket in the [parking ticket] machine,” she continued. “The first time I saw that I thought, ‘Oh my gosh, that’s so convenient.’ But it turns out that the mall has a license plate reader trained on every single parking spot. The data from that vendor who runs the parking lot then gets packaged up and sold to ICE. There’s nothing in your field of vision when you pull up that says by entering this area you are agreeing that your data can be used [in this way].”
There are compelling arguments on both sides when it comes to the use of certain controversial surveillance technologies, such as facial recognition. But for these to be debated, they must first be understood. Adversarial Fashion does its part by literally democratizing the tools it’s working with; turning a niche bug in A.I. recognition systems into something that anyone can, quite literally, try on for size.
“You don’t have to be a computer science or cybersecurity expert to use it,” she said. You can just buy a shirt. It’s very intuitive … It’s something you can physically hold in your hands. It lets you understand how these systems work. I’m not an expert who has been working on computer vision fabrics for 10 years. I’m someone with some basic functional skills and a little bit of curiosity. This is my method of how I taught myself how these systems work.”
Fashion that fights the power
One challenge, of course, is that technology continues to march on. There is no guarantee that a design that’s able to fool today’s image recognition algorithms will work with next year’s, next month’s, or, heck, tomorrow’s.
“The systems that are used for artificially identifying any kind of image get better every day,” Rose acknowledged. “The computer vision field is making constant advances. I know that my shirts may not work in a couple of years. I’ve seen old projects from other anti-surveillance artists which don’t work any more.”
In this way, it’s not necessarily any different from other politically-motivated fashion. Your “Impeach George W. Bush” hoodie is no longer so relevant in 2019, even if the broad message might remain the same.
But, hey, Kate Rose isn’t slowing down. Finding out about A.I. algorithms designed to identify humans within images (and potentially even certain types of movement), she’s now interested in creating blocky designs that, when worn on the body, can help break up the human form. One thing is for sure: Rose thinks that clothing items such as this can play a key role in the fight back against mass surveillance.
“I think one of the biggest problems we have with anti-surveillance fashion is that what is non-obvious to a computer is very obvious to people,” she said. “When you have anti-facial recognition makeup, it’s a very bold statement. You’d need to be a brave person to wear it. But with fashion you get a little bit more of a leeway. It’s not that unusual to see someone wearing a really bold print or something experimental-looking.”
It’s not going to solve the problem by itself, but it could help play a role in educating people. And maybe even causing a few headaches along the way.
“I feel it’s important to push back in any small way I can, even if I’m just one person,” Rose said.
- Inside the rapidly escalating war between deepfakes and deepfake detectors
- Facebook’s new image-recognition A.I. is trained on 1 billion Instagram photos
- An Amazon A.I. scientist wants to transform downtown Jackson, Mississippi
- Scientists are using A.I. to create artificial human genetic code
- Summon this amazing Twitter bot to colorize your black-and-white photos