London Face Recognition Identifies More Innocents Than Suspects
The live face recognition of the London police is not yet complete. Eight suspects were identified from thousands of passers-by. But seven of them were incorrectly indicated.
The Metropolitan Police, the police of the British capital, has released figures from three days on which the live facial recognition (LFR) was used: February 11, February 20 and February 27.
Passers-by were compared with a list of several thousand wanted people. The intention is that the images of passers-by are always deleted afterwards.
On that first day, 4,600 faces were scanned without a match. On February 20, the system was broken, and nobody was scanned. On February 27, 8,600 passers-by were scanned at Oxford Circus and compared with a list of 7,292 wanted people. Also, eight alerts (possible matches) came from the system.
Of those eight, two matches were incorrectly identified. Of the remaining six, five were stopped by the police, but it also turned out to be false matches. A person was correctly identified and arrested.
It is not the first time that it has become known that face recognition in London is not working flawlessly. Last summer it was announced that the University of Essex studied the system and sounded no fewer than 34 false alarms in a total of 42 matches.
The Register notes that for that reason, the Scottish government is currently refusing the technology. According to the Scottish government, the system more often gives false matches to women, blacks, Asians and ethnic minorities.
The London activist group Big Brother Watch also uses the figures to indicate that face recognition is disproportionate and inaccurate. The organization speaks of a violation of human rights and fundamental freedoms.