UK government and universities work with Chinese state institutions to advance face biometrics
A publicly-funded facial recognition research project that includes the UK Home Office and Scotland Yard does work paid for in part by the Chinese government, and has a goal of making the technology “ubiquitous,” reports The Times. Several experts and analysts are concerned that the UK government may be supporting rights violations through the collaboration.
FACER2VM is funded with £6.1 million (US$7.5 million) in grants from the UK government, and is led by academics from Surrey University. The program’s name is an acronym for “face matching for automatic identity retrieval, recognition, verification and management,” and it has published more than 100 papers since being launched in 2016. Surrey University Professor Josef Kittler is the principal investigator for FACER2VM, and says Home Office is very interested in soft biometrics and wants to upgrade its own systems.
The program could enhance the Chinese government’s ability to identify its citizens, including pro-democracy demonstrators in Hong Kong and others wearing facemasks.
Former Governor of Hong Kong and Oxford University Chancellor Lord Patten called the project “profoundly worrying.”
“It is a matter of concern when we find ourselves assisting what is a totalitarian state in developing surveillance technology used to police and incarcerate more than a million Muslims in Xinjiang. This is exactly the sort of research collaboration over which we should be extremely careful,” he told the Times.
Privacy International’s Edin Omanovic called it an outrageous use of taxpayer money.
The FACER2VM website notes that part of the technology’s value is that it can be captured without the subject’s awareness. Some research for the project has been carried out with the state-run Jiangnan University in Jiangsu, and at least a dozen papers involving FACRR2VM were financed by the Chinese and UK governments. One of those led to the development of a facial recognition algorithm with six times greater accuracy than the next best method, according to the Times, and also considered partially occluded faces.
The algorithm was successful in identifying 95 percent of people wearing sunglasses and 92.5 percent of those with a scarf covering the lower half of their faces. The next best method had success rates of 82 percent and 78.75 percent, respectively.
UK Conservative MP Bob Seely said “it’s so obvious that this could be used for nefarious purposes. We need to review our role in this project as part of a wider review of how we engage with China.”
Met officers attended a handful of FaceR2VM meetings but haa not deployed technology developed by the program, according to a spokesperson.
International Cyber Policy Centre Analyst Alex Joske said the involvement of the UK government bodies “points to a lack of effective security and due diligence.”
“It could lead to the UK government being complicit in the surveillance and human rights abuses of people in China,” he continued.
A dataset provided to UK researchers by the Home Office and other organizations for testing includes images scraped from Facebook and Twitter through Google Images.
Private facial recognition companies and the Imperial College London are among project participants, as was the BBC, which left when the direction of the program diverged from its expectation.
Kittler, winner of the Chinese government’s “friendship award” in 2016 says “society and government” are responsible for regulating facial recognition. He also called concerns about state use of research and data “unfounded,” the Times reports.
A former British diplomat to China told the Times that working with China is different than collaborating with other countries, and that China’s national intelligence laws give it potentially unfettered access to research and data.
The relationship between academics in the Western world and the Chinese government have repeatedly come under scrutiny since information started to come out about alleged mass human rights violations in Xinjiang.