Are we living in Orwell's 1984, where
computers can suddenly recognise faces? Experts reveal how developed the
technology already is and what you should be prepared for
There was a great commotion online when
Facebook released its automatic facial recognition feature in the summer of
2011. The social network could suddenly recognise faces in uploaded images and
tag them automatically. Users would help the algorithm by tagging names they
wanted to connect to faces, creating a giant database. Privacy, data protection
and civil rights campaigners became frantic. They saw the new feature not only
as a blatant violation of privacy, but a step in the direction of more pervasive
surveillance. With over 900 million active users, Facebook is the largest face
card index on the planet, and thus a goldmine for data hunters, whatever their
purpose. The feature eventually became optional but were all these fears really
justified?
Discussions about Facebook's new function
made one thing clear: there is a need for further understanding. What are the
risks involved in facial recognition, and what are the opportunities? How do
facial recognition systems work, what can they already do, and how can one
protect oneself from them?
Discussions
about Facebook's new function made one thing clear: there is a need for further
understanding
Heinrich Ihmor, a biometry expert at the
German Federal Office for Information Security (BSI), can answer these questions.
It is his task to assess the efficacy of biometric systems, including facial
recognition systems for border control and fighting against crime, in his day
to day job. At present, he finds the high-tech security gate EasyPass, which
the BSI is testing along with the federal police at Frankfurt airport, very
fascinating. EasyPass can independently conduct ID checks on passengers, find
persons wanted by the police, and compare photos using facial recognition
technology. It works well so long as travelers aren't returning from vacations
with wide, beaming smiles. "This can be thoroughly detrimental to the
biometric process", says Ihmor. If EasyPass sounds an alarm, officers
intervene and that puts a sudden end to the person's holiday joy.
The new screening method is essential
Ihmor does not, however, question the
necessity of biometric detection. Finding the best method of detection is of
the utmost importance to him. "No biometric method is perfect," he
says, "they only vary in their quality." According to him, facial
detection technology has huge advantages when compared to other biometric
methods. About two percent of the population does not have unique fingerprints.
DNA tests give outstanding results but they are expensive, time consuming and
legally disputed. Palm prints are too unclear and iris scanners are very
difficult to manage when used on a large scale. Therefore, for Ihmor, there is
no mistaking that we ought to get used to the new scanning technology.
"Facial recognition is becoming
increasingly popular", he explains. "It has made enormous progress in
the last five years." Since this technology can be used on a large scale
in a faster and safer manner, it will replace the biometric processes practiced
at present as soon as a few more technical hurdles are overcome.
"Facial
recognition is becoming increasingly popular"
Thanks to powerful high-resolution cameras
and smart software, faces can be recognised with great accuracy even from large
distances, in poor light conditions, and when they are partially covered and
distorted. The police and military benefit from this. Large areas and crowds
can be monitored with a well-positioned camera, and the systems can
differentiate between known friends and enemies. In some cases, automatic
systems can open doors or even deny someone access to a computer terminal, for
example.
The list of fields of application is
growing longer. For a long time now, the American military has been using
drones as people-hunters: thanks to facial recognition technology, they trace
targets in no-man's-land even from great heights without these people noticing
anything. At the next football world cup, the Brazilian police force intends to
equip officers with special goggles with integrated cameras which can flag
notorious hooligans and make them stand out more in their fields of vision.
Apparently these goggles can process 400
faces per second and compare them with a comprehensive database that can
contain up to 13 million faces. Local transportation services in Rotterdam intend
to use facial recognition technology to implement bans on people entering
public vehicles; a guard can automatically be called if a known troublemaker
enters a bus or train.
New algorithms can not only compare faces
but also analyze them. What is the mood of the person, where is he/ she
looking, how old would the person be? This is the kind of information that even
those in the advertising industry are desperate to get their hands on. It can
help in analysing the influx of visitors in a supermarket or shopping
disctricts, and display ads on intelligent billboards that are tailor-made to
the gender and age of passers-by.
Will face scanners replace door locks?
So far, facial recognition technology has
been most widely used in the entertainment industry. Image programs such as
iPhoto and Picasa can sort large photo collections on the basis of faces.
Portable cameras can snap photos when all the people in the frame are smiling,
and smartphone apps help in recognising faces of people we might have met long
ago. Gaming consoles detect players by face and the technology is replacing the
password input in an increasing number of protected areas, right from websites
to door locks.
Will
face scanners replace door locks?
Yet, such popular examples paint an
inaccurate picture of the current applications of facial recognition
technology. BSI expert Heinrich Ihmor points out: "If I know all about the
limits of biometrics, I tend to overrate its results." For Ihmor, facial
recognition technology is basically "a wonderful tool for providing
technical and hassle-free support to those who are involved with
detection."
This is why one must not rule out human
beings as a source of errors. "Biometrics depends on enrollment and the learning
procedure," explains Ihmor. "This is time consuming and requires a
lot of work." Therefore, he will not trade the lock on his house door for
a face scanner - and because he doesn't want to be locked out of home, for
example after coming back from a vacation.
The German Federal Criminal Police Office
must have understood that facial recognition technology still needs
improvement, when it tested three camera systems for monitoring the entry area
of the Mainz Central Station between October 2006 and January 2007. The 23,000
daily visitors also included 200 pre-selected test subjects. The result was
sobering: even under ideal conditions, the systems caught only 60 percent of
the times they passed by. The recognition rate fell to between 10 and 20 percent
in poor light conditions. There were 23 false alarms every day, on average.
This was especially annoying for security personnel as well as those disturbed
by the alarms. The final report stated that using the system in open air is
currently "not promising" for manhunts. Despite many discussions,
there is still no method devised to identify hooligans in football stadiums, a
potentially useful application.
On the other hand, the suggestions of BKA
experts to increase the recognition rate by 'conscious or unconscious
cooperative behavior of the wanted person" seem to be curious: a moving
screen next to the camera coaxes the wanted persons to raise their heads - a
trick, for which hardly any real criminal or terrorist would fall.
Despite the setback, the automatic facial
recognition technology has been playing an important role in the daily routine
of the German investigator for a long time. "Basically, the facial
recognition technology has become an indispensable method of person
identification," a BKA speaker explained to us. As a matter of routine,
officers compare the photos of the offenders with the portraits saved in the
central information system of the police (Inpol) according to the information
provided by BKA, the system contains 3.3 million photos of persons "who
have been subjected to the recognition service" persons. However, even
here the photos cannot be compared without human help. The decision is to
whether a hit is the same as the wanted person, is always taken by the experts.
The computer only presents a list of photos sorted on the basis of similarity.
