| Face reconition software works for fighting terrorism |
| BKA in Germany conducted a large field test to see if face reconition software could be used to catch criminals and terrorist suspects. |
| While the media got it wrong including the BKA’s own president Jörg Ziercke – from a scientific perspective – these tests were a success |
| With minimal human interference – false positive ratio was kept quite small – 0.1 percent |
| Face recognition technology if used correctly and according to the country’s constitution and privacy laws is an effective tool for catching criminals and stopping hooligans from entering a football stadium. |
The BKA (German equivalent of the FBI) tested face recognition technology using the main railway station in Mainz to conduct the field test.
p. 27 states:
- Durch einfache Massnahmen sollte es moeglich sein, Einfluesse die sich im hiesigen Feldtest noch negativ auf die Erkennungsleistungen ausgewirkt haben, zu minimieren. Durch massnahmen, die die Blickrichtung der Probanden auf die Kamerapositionen lengen (z.B. eine Laufschrift in der Naehe der kameras) waere es ohne weiteres moeglich gewesen, nochmals bessere Vergleichsbilder aus dem Videostream zu erhalten. Eine ensprechend hoehere Anzahl an Frontalaufnahmen haette zweifellos die Anzahl an richtigen Erkennungen steigen und die Anzahl falscher Erkennung sinken lassen.
The above could be translated as saying that simple measures that make sure people are guided in passing through security controls (e.g., football stadium -editor) suffice to assure high recognition ratios. For instance, getting the person to look toward the area where the cameras are located (e.g., having a video ad/film run nearby where the camera is mounted to get people to look in that direction) would have improved the study’s results significantly. In fact, a higher number of front view pictures would have reduced the false-positive and the false-negative ratios obtained by the study even more. The report continues:
- Wenn die Moeglichkeit gegeben ist, die Menschenstroeme zu vereinzeln und ein kooperatives Verhalten zu bewirken, beispielsweise im Rahmen einer Einlasskontrolle, steht zu erwarten, dass ein Grossteil der gesuchten Personen von einem biometrischen Gesichtserkennungssystem zuverlaessig erkannt wird.
The above means that when the possibility is available that massess of people can be guided to support the face recognition efforts (which was not done in Mainz – the editor), such as in the case of entry control, one can expect that the majority of people being searched for will be correctly identified by this technology.
- Eine realistisch annehmbare Falschakzeptanz-/Verwechslungsrate von 0,1% ist ein handhabbarer Wert. Sie zeigt aber auch, dass die letzte Entscheidung darueber, ob Personengleichheit besteht, immer von einem Menschen getroffen werden muss, der die angezeigte Erkennung bereits am Bildschirm beurteilt. Am Hauptbahnhof Mainz waeren ansonsten taeglich ca 23 Buerger aufgrund von Verwechslungen mit weiterfuehrenden Massnahmen belastet worden.
A realistic and acceptable false-positive or false-negative ration of 0.1 percent is a workable number. It also illustrate that the last decision to be made if an identity exists between the searched for person and the one passing through the security control must, however, be made by a human making this decision when comparing images on the screen. Else the field study at the main train station in Mainz would have resulted in 23 errors made each day, whereby citizens would have been checked further due to a mistake caused by the face-recognition technology.
WHAT IS NOT CORRECT
The claim by some media that the BKA basically declared the field-study worthless in terms of being an investigative tool is NOT correct (see above facts).
It could be true – if the reporter did not misquote – that apparently the president of the Bundeskriminalamt (BKA) Jörg Ziercke may have said something to the fact that he perceived the study as not being a success …. but more likely the journalist got it wrong and took the quote out of context, see here:
What happened – the Spiegel reporter failed to read the report properly
FACT IS — AS STATED IN STUDY
Most false positives (estimated to be 23 a day in the field test undertaken at Mainz main train station) can be eliminated if the final decision – is the match really a match – is being made by a human viewing the pictures on a computer screen before launching the alarm.
Data protection and privacy officers are against it…. —- the study does not mention this at all. However, German press quoted some unnamed privacy commissioners of being against this method if it is used in public places. Their fear is that so-called dragnet methods are applied to find criminals in public places. However, Germany’s consitutional court in Karlsruhe has already said that the drag-type method for finding criminals in public places (in contrast to the UK) is not allowed (see further below for link to story).
Get the full report about the BKA face-recognition field-study here:
CONCLUSION
The study demonstrates that the technology can be used nicely and is actually very suitable for surveillance and facial recognition when trying to match possible suspects in a manhunt.
How the technology is applied and if laws allow its application under what conditions is another issue. Germany already has had a court case where using video cameras in a public location for security purposes was ruled to be illegal since it violates people’s rights according to the countr’s constitution:
Of interest is also the trend to archive more and more communication data in the hopes of finding terrorists and criminals. Using face-recognition technology in combination with telecommunication data (e.g., voice, location as collected from user’s mobile phone) is one step to a state whereby the individual’s rights for privacy are surely no longer protected, are they?
- Regulation that matters – Spain PROYECTO DE LEY 121/000128
ADDENDUM
And no, the face-recognition field test was not intended to confirm or disconfirm the technology’s applicability for identitifying millions of people all accross the country. Instead, it was a field test in a train station and, most importantly, the results show that with the right set-up, results can be encouraging. Now let us just reflect a bit about privacy issues and move beyond the important issue of false positive or negative ratios we get with using this technology.
SUBSCRIPTION
To make it more convenient for you to take advantage of CyTRAP Labs’ offerings, just provide us with your e-mail address below. You can personalize your subscription to make it suit your needs.