Discussion Prompt: Should we ban the use of automated video-surveillance? See all contributions to this question.
Automated facial recognition is much better than its reputation would suggest and has a lot to offer as a tool for enhancing public security. There is no reason why the technology should not be used to help prevent and detect crime at known hotspots, so long as rigorous constitutional and operational standards are met.
The use of automated facial recognition techniques would be a major win for German public security, yet such use is currently heavily condemned by its critics. There is talk of it as the “dictator’s instrument” and of “total state surveillance”, with sometimes even Orwellian dystopias being conjured up in which the German state observes all of its citizens at all times in order to correct and educate them. All this — it must be clearly stated — is utter nonsense. This article will explain how such scenarios have nothing, absolutely nothing, to do with what is being discussed in Germany with regard to automated facial recognition and what is being advocated by the CDU and CSU.
Constitutional and effective
Under German constitutional law, it would not even be permissible to use automated facial recognition on a broad scale. Accordingly, it can and is intended to be used exclusively at crime hotspots and critical security facilities, such as airports or large train stations which have been placed in the highest risk category by the Federal Police. Nobody is intending to use facial recognition technology in all of the more than 5,500 train stations around the country, which are generally unlikely to be targeted by terrorists. We are therefore only talking about a selective collection tool to be used at a limited number of places where there is a special interest in their security.
The matching of the video feed should only be carried out with a clearly defined reference database, i.e. a narrowly-defined and limited search base. The facial recognition system would therefore only sound the alarm in the case of wanted felons or terrorists. In all other cases, the program would not be able to identify a person and would immediately delete the corresponding data. And there is no reason why this program should not be evaluated by the Federal Commissioner for Data Protection and Freedom of Information (BfDI). The anonymity of citizens in public places is therefore not threatened at all. Even in the case of a match, the decision to carry out police measures would be made without exception by the police officers on site after reviewing the results and evaluating all relevant information and circumstances.
Contrary to what is claimed by the opposition parties in parliament, automated facial recognition is not a half-baked technology. It is true that, similar to the human ability to recognise faces, the recognition accuracy of automated systems is negatively influenced by numerous factors. These include, for example, poor lighting conditions, partial masking of the face, unfavourable perspectives, different camera optics, and different camera resolutions. However, tests conducted by the German Federal Police from August 1, 2017, to July 31, 2018, at Berlin-Südkreuz station on intelligent video analysis using biometric facial recognition were very successful and have already demonstrated that, when two different systems are combined, the rate of false matches drops to a negligible 0.00018 percent.
With a false-match rate of 0.00018 percent, there would only be two false matches per one million. For comparison: The daily passenger volume at a large central station such as Frankfurt or Berlin is between 360,000 and 500,000 people. Before the outbreak of the pandemic, Frankfurt Airport recorded around seven million passengers per month during the peak travel season. That means that even at heavily frequented transit hubs the absolute error count would be quite low. Whether two images show the same person can be determined much more accurately by facial recognition technology than even by professional image evaluators, known as ‘super recognisers’, not to mention the false match rate of a human police patrol, which can be assumed to be much higher.
An overview of the criticism
It is often claimed that the technology of automated facial recognition does not prevent crime. I cannot understand this claim, however, because with the same logic one could also argue against the criminal code. A punishment imposed by a court also failed to prevent the crime in question. But of course it still has a general preventive effect. The same is true of video technology, for no criminal is keen on clear images of his or her crime being available.
It is often claimed that a majority of the population rejects intelligent video technology. As far as I can see, there is only one 2019 survey that suggests this, and in this survey a majority is actually in favour of its use under certain conditions. Only 22 percent of those surveyed were in favor of a total ban, whilst 17 percent were even in favor of the unrestricted use of the technology. In all surveys (e.g. this one by Forsa and another one by YouGov) on the use of classic video surveillance, a very clear majority of respondents even call for its expansion, and it is particularly women who support its use and expect this technology to provide greater security.
It is often said: We do not need more video surveillance but more police. This is true, but it conjures up a false choice – as if you can only have one or the other. During the last legislative term, the grand coalition decided to create 10,000 more jobs in the police and security agencies of the federal government and will have created another 7,500 jobs at the federal level alone by 2021, by means of the Pakt für den Rechtsstaat (Pact for the Rule of Law), plus 7,500 more in the German federal states.
A step up for our security
Regarding the constitutional evaluation of facial recognition by law enforcement agencies, it must be said that its use is certainly a delicate subject with regards to basic rights, but that such use is constitutionally possible if high standards are observed. It is part of the basic freedom we enjoy in our democracies that citizens can generally move around without being arbitrarily registered by the state, having to account for their integrity, or having to worry about constant surveillance. Nobody wants to restrict this freedom. The measures within the framework of police facial recognition must therefore target crimes of a grave nature, be limited to crime hotspots and, of course, high standards must be observed to protect against discrimination. This means that the emergence of discriminatory patterns within the inevitable errors must be prevented.
Expanding the possibilities for facial recognition in public spaces makes an important contribution to preventing and solving crimes. Based on the results and experiences from the project at Berlin Südkreuz station, the use of biometric facial recognition could make the fight against terrorism and serious crime more successful. Using the technology in Germany would represent a major step up in our security policy.