Feeds

Reverse-engineering artist busts face detection tech

Hacking Big Brother with help from Revlon

Next gen security for virtualised datacentres

Concerned about the proliferation of face recognition systems in public places, a grad student in New York is developing privacy-enhancing hacks designed to thwart the futuristic surveillance technology.

Using off-the-shelf makeup and accessories such as glasses, veils, and artificial hair, Adam Harvey's master's thesis combines hipster fashion aesthetics with hardcore reverse engineering of face detection software. The goal: to give individuals a low-cost and visually stimulating means to prevent their likenesses from being detected and cataloged by face-recognition monitors.

"The number of sensors that are going into the public spaces has been increasing," said Harvey, a student in New York University's interactive telecommunications program. "There's a lot of work to be done to catch up to where cameras are going because there have been so many advances in the last few years."

Although still in its adolescence, face recognition technology is quickly being adopted by governments and corporations to identify individuals whose images are captured by surveillance cameras. At the 2001 Super Bowl, for instance, officials digitized the faces of everyone entering Raymond James Stadium in Tampa, Florida and compared the results against photographic lists of known malefactors.

In another example, the city of Chicago two years ago made much fanfare of "Operation Virtual Shield," which was to use IBM software to analyze in real time thousands of hours of video being recorded on more than 1,000 continuously running cameras.

As a starting point, Harvey's research involves the reverse engineering of OpenCV, which its creators describe as an open-source "library of programming functions for real-time computer vision." From that work, he developed an understanding of the algorithm used to tell if an image captured by a camera is, say, a car, a building or a human face.

Based on the so-called Viola-Jones method (pdf), the algorithm examines the spatial relationships of an object captured in an image and looks for features commonly found in faces. Most faces have a dark region just above the eyes, while the cheek bones and nose bridge will appear lighter. When the algorithm detects enough such attributes, it guesses the object is a face. The method is generally regarded as effective. Errors are in favor of false positives, making it hard for unobstructed faces to escape notice when they aren't captured at an angle.

Once a face is detected, other technologies, such as face recognition, can be used to compare the face against a database in an attempt to identify the person it belongs to.

But Harvey has discovered that face detection can often be thrown off by using makeup to alter the contrasts the technology looks for. For example, dark patterns applied around eyes and cheek bones, as in the image below, are one such possibility.

These faces aren't detected by Viola-Jones algorithms

"There's a lot of trial and error," Harvey said. "The common thread is throwing off the symmetry" the algorithm looks for. "It's a lot more difficult than applying a bunch of makeup and hoping it works or putting on your 3D glasses left over from Avatar."

Technology that detects or recognizes faces still hasn't gone mainstream, said Ralph Gross, an expert in the field and a scientist at Hyperactive Technologies. For now, it's mostly limited to specialized applications, such as the Super Bowl outings or Las Vegas casinos, but he said he expects that to change.

"The technology is getting better all the time," he said. "Along with computing power being greater, it becomes more of an option. I think we're heading to the point where we as a society need to think about what we are comfortable with."

At the same time, Gross said a pair of dark sunglasses or a simple veil would probably prove just as effective at thwarting face detection as anything Harvey is recommending.

But Harvey, said the point of his project from the beginning has been to create disguises that do more than simply hide a person's face.

"The combination of hair, makeup and accessories gives you the potential to do an infinite number of creative new looks that have some futuristic value to them with anti-face-detection functionality," he said. "Maybe you could go to a privacy hair stylist in the future." ®

The essential guide to IT transformation

More from The Register

next story
Goog says patch⁵⁰ your Chrome
64-bit browser loads cat vids FIFTEEN PERCENT faster!
Chinese hackers spied on investigators of Flight MH370 - report
Classified data on flight's disappearance pinched
NIST to sysadmins: clean up your SSH mess
Too many keys, too badly managed
Scratched PC-dispatch patch patched, hatched in batch rematch
Windows security update fixed after triggering blue screens (and screams) of death
Researchers camouflage haxxor traps with fake application traffic
Honeypots sweetened to resemble actual workloads, complete with 'secure' logins
Attack flogged through shiny-clicky social media buttons
66,000 users popped by malicious Flash fudging add-on
prev story

Whitepapers

Top 10 endpoint backup mistakes
Avoid the ten endpoint backup mistakes to ensure that your critical corporate data is protected and end user productivity is improved.
Implementing global e-invoicing with guaranteed legal certainty
Explaining the role local tax compliance plays in successful supply chain management and e-business and how leading global brands are addressing this.
Backing up distributed data
Eliminating the redundant use of bandwidth and storage capacity and application consolidation in the modern data center.
The essential guide to IT transformation
ServiceNow discusses three IT transformations that can help CIOs automate IT services to transform IT and the enterprise
Next gen security for virtualised datacentres
Legacy security solutions are inefficient due to the architectural differences between physical and virtual environments.