Matthias Monroy

Since 2008, German police authorities have been using a facial recognition system to identify unknown persons. The facial database queried in this way has grown dramatically in the past year.

The German Federal Police more than doubled the number of hits when searching for persons using facial recognition last year. In 2022, 2,853 unknown persons were identified using the technology, compared to 1,334 in the previous year. The number of searches also increased in the same period, but to a much lesser extent. In 2022, the Federal Police conducted 7,697 searches, compared to 6,181 the year before.

The figures come from the answer to a parliamentary question by the Left Party, in which it asks for figures for biometrics-based applications every year. Therein, the Federal Ministry of the Interior was also asked to provide data for the federal as well as state criminal police offices. However, these are not yet available, according to the ministry.

Facial images of 4.6 million persons

The Federal Police uses facial recognition to investigate criminal offences. For this, they use mobile phone photos of suspected perpetrators or recordings from video cameras in public places.

The searches are then carried out using the facial recognition system of the Federal Criminal Police Office (BKA), which has been available to all German police authorities since 2008. In the process, biometric photos are queried against the INPOL file.

This largest German police database is also maintained centrally by the BKA for all police forces. It contains photos from identification treatments of suspects as well as asylum seekers. In addition to the INPOL file, the BKA also has a file „ST-Libi“ with currently 3,571 searchable photos „for state protection purposes“.

The number of photographs in INPOL has increased noticeably in the past year. In 2021, around 5.5 million portrait images of 3.6 million people were stored there in a searchable form. When answering the parliamentary question, there were about 6.7 million images of 4.6 million persons.

The BMI also mentions the reason for the significant increase: in 2022, almost 1.5 million images were added, but only about 400,000 deleted. Such deletion happens after the end of the storage period of an individual entry. However, the ministry does not explain the unusually high difference between newly stored and deleted persons. Last year, the number of images had dropped by about 400,000.

„Dirty data network“

In the future, police facial recognition will probably increase. According to the Ministry’s answer, the Federal Police is currently testing two systems from the companies Digivod and Idemia Germany for „semi-automated video analysis“. This involves the processing of so-called mass data, i.e. recordings from public spaces in which people’s faces are recognisable.

Both companies have previously participated in research projects for the analysis of „image and video mass data“ from the Federal Police or BKA. In the PERFORMANCE project, the participants perfected an upload platform like the one the police made public for the submission of photos and videos after the New Year’s Eve in Cologne in 2015 and the G20 summit in 2017, when the police searched afterwards for a big number of offenders.

In FLORIDA, in addition to visual analysis, the recognition of patterns in sound recordings was also tested. The company Idemia was also involved in the pilot project of the Federal Police and Deutsche Bahn for real-time facial recognition at Berlin’s Südkreuz station.

The Federal Police’s tests on „partially automated video evaluation“ are taking place within the framework of a newly founded „AG Digitale Kompetenz“ (Digital Competence Working Group), which is overseeing, among other things, the development of a „dirty data network“. This is a data repository, on which the publicly collected mass data is stored and processed, is to be kept separate from the actual police data grid.

Image: Test of facial recognition at Südkreuz/ Berlin (Federal Police).