Fool Facial Recognition Technology

Email Print
FacebookTwitterShare

 

 
 

Visual Recognition
technology is becoming very powerful and easily has the capacity
to perform facial recognition of an entire crowd. They can probably
see you reading the How
To Vanish
book from space so scanning a crowd is an easy
feat. Facial biometrics can link the faces in an otherwise anonymous
crowd with the vast transactional
databases
that are maintained on almost everyone. Your physical
presence in a particular location, detected by face recognition
technology, would be yet another data point in the vast digital
dossier on you. Such visual recognition can pose serious threats
if misused to anyone who wants to participate
in the political process
, peaceably
assemble
, participate in religious observance or in many other
aspects of life. So I thought you might want to know how to fool
facial recognition technology.

Visual Recognition
Technology

The fact that
facial biometric technology exists is not the problem. It is the
use of technology, such as visual recognition, that can have serious
implications. Use of face recognition technology is not new. It’s
use was first made public several years ago when it was used to
scan the crowd of the Super Bowl in Tampa Florida. At that time
face recognition was not very good and returned a lot of false positives,
identifying individuals as someone they weren’t, as well as
missing other people the system was designed to detect. Such a deployment
of visual recognition in its unrefined state may have caused the
public to regard the technology as harmless. Since that first deployment,
the technology has improved and is now used for airport security
and in other places where there are a lot of people to spy on and
very little time.

Fool Facial
Recognition Technology

Adam Harvey
undertook a very interesting project as part of his doctoral thesis
at NYU to find ways to fool facial recognition. Makeup was applied
to faces in these patterns to obscure facial features. The researcher
relied on the algorithms that the software used to recognize faces.
Knowing the algorithm allowed him to manipulate and distort the
most important reference points to avoid detection by the facial
biometrics.

Limitations
of Facial Recognition

Keep in mind
that this is the first study of its kind so there are some limitations
in what is now known. For example, the software does not even recognize
that a face is present at all when using these patterns. The study
does not address whether there are ways to allow the software to
recognize a face without being able to determine which face it is.
The contrast that must be used is also not reported. Do the patterns
require the use of black and white makeup, or can we use the colors
of our favorite sports teams? Are more subtle shades useful at avoiding
detection that might not draw as much attention as a face painted
like Gene Simmons? Hopefully others will follow up and find the
answers to these questions.

Implementing
Visual Recognition Disruption

For now, if
you follow the patterns as presented, it is likely that your face
will not even be recognized by facial recognition software. I would
not recommend testing new techniques or even using proven techniques
in any heavily regulated areas, such as airports. Even Gene Simmons
might have difficulty getting through airport security with face
paint. But who is to keep you out of the Superbowl with Chargers
blue and gold displayed proudly? Even if the Chargers aren’t
playing.

Conclusion

Visual recognition
technology is advancing quickly and becoming a very powerful tool
to monitor a higher number of people, more often, at a lower cost.
Keep your facial biometrics to yourself and stay out of the transactional
databases
, avoid
surveillance cameras
more effectively, and keep the paparazzi
guessing. If there have been any interferences with your constitutional
rights due to the use of facial recognition technology, you will
want to talk to a lawyer.

Reprinted
with permission from How to
Vanish.

November
30, 2010

Email Print
FacebookTwitterShare
  • LRC Blog

  • LRC Podcasts