The Federal Criminal Police Office can search a database with almost 6 million facial images, the system is now being equipped with artificial intelligence
Last week, the German Federal Minister of the Interior Horst Seehofer surprisingly moved away from plans to expand facial recognition in public spaces. He had demanded that the use of so-called intelligent video surveillance be anchored in the Federal Police Law. In the current draft of the new law, the topic is now excluded. However, it is questionable whether this really means a renunciation of the technology. The Ministry is of the opinion that § 27 of the Federal Police Act allows the automatic evaluation of camera images anyway. It states that the Federal Police may “use automatic image recording and image capturing devices”. Actually, this meant automatic continuous operation and remote control of video cameras. In the legal literature, it is therefore disputed whether the analysis of images using algorithms or artificial intelligence is covered by this.
If video surveillance is to be introduced in real time and in public spaces, it must be decided which files the cameras access. It can be assumed that this will be applied first in areas that meet with public consent. These include the search for missing children or the search for crimes against sexual self-determination. It is also conceivable to have the cameras search for so-called dangerous persons or all persons with open arrest warrants.
Many of the persons concerned are already stored with their facial images in the INPOL file, the police data network of the German federal and state governments. This is operated centrally by the Federal Criminal Police Office (BKA) and now contains around 5.8 million photographs. Most of them come from police ID treatments, which are carried out after criminal offences or as a preventive measure. Usually fingerprints and a facial image are taken. However, the file probably also contains photos of migrants who are required to leave the country. The drastic increase in the number of stored data is inexplicable. In 2016 there were still 4.86 million photographs of 3.34 million people, so in four years around one million more pictures were added. In the same time, however, many photos would have had to be removed due to deletion deadlines.
Since 2008, the BKA has been using a face recognition system from the Dresden-based company Cognitech to search the image collection in INPOL. Investigations are carried out, for example, when a high-resolution photo of an unknown person is available after a crime. The system also recognizes when a person is stored under multiple identities. In addition to the BKA, the state criminal investigation offices and the federal police also access the system. The use of this retrograde face recognition is also increasing significantly. While in 2016 there were still around 23,000 queries, these have now roughly doubled. In 2018, around 1,000 people were identified in this way, but final figures for 2019 are not yet available.
The BKA now wants to modernize its face recognition system. In earlier replies to parliamentary inquiries, the Ministry of the Interior had already stated that ears would also be used to measure the face. For corresponding tests, the BKA had procured systems from various companies, including Cognitec, AnyVision and ldemia, the market leader for European biometric systems. The BKA intends to award the contract for the new system to one of the suppliers before the end of March. As with the governmental trojan software, the authority operates blindly with facial recognition. The “exact functioning of the systems” is not known to the BKA, writes Hans-Georg Engelke, State Secretary in the Federal Ministry of the Interior. The new facial recognition might also use “methods of machine learning”. This hopefully calls for the Federal Data Protection Commissioner, who, as with the State Trojan, can demand insight into the source code of the software.
Image: Action against video surveillance at Südkreuz station in Berlin (
).
Leave a Reply