Masked protests and “black bloc” direct actions could soon face a new danger from the national security state.
But this time, the threat won't come from infiltrators or informants. It will come from an artificial intelligence algorithm capable of identifying targets based on imagery revealing only a small part of a person's face.
The new deep-learning algorithm developed by researchers from Cambridge University, India's Institute of Science and India's National Institute of Technology connects points from a person's face using a “star-net structure” and uses the angles to identify their unique features, regardless of whether the targeted individual is wearing a partial disguise, scarf, mask, cap or sunglasses.
The new research may impact the ability of social movements and protesters to hold mass mobilizations in conditions of repression, where participation in unauthorized assemblies could lead to punitive repercussions, such as jail time or worse.
Cambridge Ph.D student Amarjot Singh told Motherboard, however, that the project was originally meant to identify criminals. He remains unsure of how the technology can be prevented from falling into the hands of governments inclined toward using the tool to persecute social movements.
"I actually don't have a good answer for how that can be stopped," he said. "It has to be regulated somehow … it should only be used for people who want to use it for good stuff."
Despite the researchers' hopes, the industry of repression thrives on marketing such high-tech, tailor-made solutions to political dissent to authorities across the globe in countries like Myanmar, Saudi Arabia, the United Arab Emirates and Azerbaijan.
A photo from the research paper shows how facial key-points are used to perform face identification. | Photo: arXiv.org
“Unfortunately, the American government, the world’s most vociferous defender of 'Internet freedom,' has little to say about such complicity,” tech critic Evgeny Morozov wrote in 2011 during the height of the Arab Spring. “Such reticence may not be entirely accidental, since many of these tools were first developed for Western law enforcement and intelligence agencies.”
“Western policy makers are therefore in a delicate spot,” he added. “On the one hand, it is hard to rein in the very companies they have nurtured; it is also hard to resist the argument from repressive regimes that they need such technologies to monitor extremists. On the other hand, it’s getting harder to ignore the fact that extremists aren’t the only ones under surveillance.”
Luckily, the program isn't quite ready for prime time, but the research provides crucial groundwork for future breakthroughs that can sufficiently train the algorithm to achieve a near-perfect identification score.
And while the cliched “Anonymous” Guy Fawkes masks or presidential masks made famous in the film “Point Break” may shield one's entire face from facial recognition programs, research is being undertaken that can identify people based on how they walk by analyzing the unique gait signatures of individuals using advanced machine learning algorithms.
Let's hope that surveillance technology doesn't develop to the point where we have to roll out to future protests in wheelchairs while wearing sweat-inducing latex Ronald Reagan masks in order to avoid the all-seeing eye of Johnny law.