Make-Up tricks for hindering recognition
The biggest technical problems will however
be solved soon. Researchers have been working for a long time on scanners which
combine various biometric methods: a 3D scanner tries to determine the line of
vision and inclination of the subject's head to correct the image of a 2D
scanner. A face scanner checks the identity, an iris scanner (which functions
from a distance of up to two meters) counterchecks it, and a thermal image
camera decides whether it is a real face or a rubber mask. "We can improve
facial recognition technology dramatically by combining various methods",
says an excited Ihmor.
And what will happen next? Total pervasive
surveillance? There is hardly any surveillance tool that provokes such a strong
backlash as automatic person detection. Artist and civil rights campaigner Adam
Harvey is not surprised: "The problem with facial recognition systems is
the imbalance of power" he says. You can't tell whether a camera just
passively records video or analyses what it has seen. On the other hand, complete
surveillance infrastructures can be set up using relatively simple software
updates - think of the million or more cameras in Great Britain. Today, those
who resist a scan arouse suspicion. Harvey has found a subversive workaround:
make-up tips that can outsmart facial recognition systems. A streak of hair
here, a black stripe there, and the computer can no longer recognise you. His
project "CV Dazzle" was an Internet sensation - and shows yet again
that the topic of facial recognition must be discussed on priority.
Does Facebook serve as a police archive?
This has been going on in politics for
years now, for example in the office for technology assessment of the national
parliament. A report from the year 2002, whose point of view is applicable even
today, says that only a public, transparent discourse can "create
awareness about the significance of the dynamics of the social-technical
development [...] which may be combined with intensive application of
biometrics in future." In short, biometric systems are a serious
disruptive force in the balance of control mechanisms of today's societies.
Artist
Adam Harvey's project "C V Dazzle" showed how deliberate alterations
to a person's appearance could defeat face recognition systems.
This affects everyone, because such systems
are practical only if as many relevant persons as possible are recorded. This
is also the reason why the Facebook venture shocked data protectionists and
civil rights campaigners: not only do they fear the actual potential for surveillance,
but also the changed mentality that will emerge. Once you understand the
convenience of biometrics on Facebook, Picasa or Google, you will probably no
longer be scared of its darker side and become careless when dealing with
unknown identities. When asked about this phenomenon, data protection officer
Peter Schaar expressed his concern: "I find it unsettling that any random
person can be identified with the help of digital photos which have been taken
by a smartphone, by comparing them with the photos that have been published on
the Internet or on social networking sites."
But at least one concern is unfounded at
present: that police authorities can access these private photos saved on the
Internet to arrest offenders through mass face recognition. In many countries,
biometric features are considered to be personal data that is exceptionally
worthy of protection. Involuntary imposition, storage and processing are
permitted only in exceptional cases and the person concerned must be informed
of this. Even if it were easier to access Facebook's index of saved faces,
there are very practical reasons that prevent the routine use of such a
feature. A policeman explained: "We simply don't have the manpower for
that." That may well be true, but it will only last as long as computers
still need human support.
New methods
There are many facial recognition methods,
but none of them has been accepted as the right one. However, the world is
clearly moving from simple geometrical processes to combo solutions.
2D geometry
2D
geometry
In the initial stages of facial
recognition, simple geometrical data was compared: the distance between the
eyes or the distance between the chin and hairline. But this method is too
inaccurate and too easy to outsmart.
2D statistics
2D
statistics
More accurate and statistical methods such
as the Eigenface approach are used (right). Here, an abstract image is
calculated and the differences between the test face and the reference face are
checked. Another approach compares the frequency range of test faces using
Fourier transformation.
2.5D combo
2.5D
combo
The statistical methods can be enhanced
with 3D information, for example correcting the line of sight of a test face.
This eliminates image errors or deceptions such as holding up a photo of a face
to the sensor.
3D recognition
3D
recognition
Laser scanners or sample projections (such
as those in Xbox Kinect) are especially difficult to outwit, and are at least
theoretically independent of lighting and facial alignment. But they are costly
and susceptible to discrepancies such as caps or lifted collars.
Combinations
Combo solutions provide the best
recognition data. These solutions make use of other biometric methods, for
example iris checks (from up to 2m away), finger vein scans, face thermal
imaging, lip movement, gait recognition and/or keystroke patterns.
Be anonymous on Facebook
Be
anonymous on Facebook
It's easy enough to make sure that Facebook
does not automatically try to analyse your face:
Deactivate Facial Recognition: Open the
"Privacy settings" in the drop-down menu at the top right (small
triangle). Then go to "Timeline and Tagging | Edit settings | Who sees tag
suggestions when photos that look like you are uploaded?". In the next
pop-up window, "Tag suggestions", select the "No one"
option.
Delete Saved Data: Facebook stores summary
data about your photos in order to detect your face in others' photos and
suggest tags. The site will delete this information upon request, but it's
difficult to actually do so. You will have to contact Facebook yourself and
request that they delete your summary data - an automated form that allowed you
to do this yourself seems to have been removed.
Test a facial recognition system
You can use facial recognition systems
yourself. Several PC utilties and smartphone apps are available.
KLIK by Face.com: If you turn your iPhone
camera towards a friend, this app displays his/her name. For this, the app
searches biometric data and the names of your Facebook friends. With just a
click, you can also teach this app new names and faces.
Test
a facial recognition system
Avatar Kinect/Windows 8: Microsoft's Xbox
360 can also detect faces with the Kinect add-on and automatically select
suitable chat avatars. Microsoft intends to integrate this technology in
Windows 8 as well: if a new person sits in front of it, the computer can
automatically change the user account.
Android 4.0: Instead of a code lock, the
latest version of the smartphone OS can also be protected via facial
recognition. But be careful; the "Face unlock" function can be
manipulated by holding up a photo of the owner.
