(Forbes) – The Russians are pretty good at spear-phishing (ahem) and “highly persuasive, election altering” social media advertising, but they are even better at developing facial recognition systems. So great, in fact, that the National Institute of Standards and Technology (NIST), part of the Commerce Department, awarded Moscow-based NTechLab with this year’s Face Recognition Challenge Prize.
The first-ever facial recognition competition was devised the Intelligence Advanced Research Projects Activity (IARPA), the research and development team run by the Director of National Intelligence. NIST scientists were the ones who chose NTechLab for their algorithms.
“Our job is to be the best and to build the best technology in the world you have to work with the best in the world,” says IARPA’s program manager for the facial recognition prize, Christopher Boehnen. “This is like a Nascar race. There are three different categories of prizes and NTech won two of them. It doesn’t matter where you are from. So long as your country has a relationship with the U.S., you’re welcome to try your luck.” Boehnen was one of the judges.
NTechLab beat out 16 companies.
They won for best one-to-one verification, which would be like using your face to log into your phone. And they won on speed and accuracy. China’s YITU won for identification accuracy. NTechLab placed second in that category. Other competitors included U.S. based Rank One Computing; Imperial College London; Digital Barriers of the U.K.; Ayonix from Japan; Morpho from France, and Deep Sense of China.
Winners got a plaque recognizing their achievement and a check for $25,000.
With recognition comes the potential for procurement contracts.
“We’ve spoken with at least twenty companies since winning this prize,” says NTechLab’s 27-year-old co-founder Artem Kuharenko. “There’s been a lot of requests from new clients from all over the world, including American companies,” he says.
Kuharenko’s NTechLab is best known for their FindFace.ru app on the VKontakte social network in Russia. The two-year-old company has around 50 employees, with most of their engineers based in Moscow. NTechLab’s marketing team is based in New York.
NTEchLab sent its FindFace algorithm to the competition. The image dataset had 500 million faces with a reported accuracy of over 80%. NTechLab runs a project for an undisclosed multinational company with one billion faces for searching. Their technology can be used to identify people in real time through closed-circuit video security systems. Moscow uses it, as Bloomberg reported in September, to “spy on its citizens in (the) streets.” (Perhaps our street cams just take candid pictures and send them back to people should they wish to buy them.)
One way NIST judges deciphered who had the better algorithm was to take photos of the same people in different lighting and angles and pump them through the system of roughly two million images to see which algo finds them first, and which ones were more right than wrong.
“It takes less than one second to find you,” says Kuharenko from his home in Moscow. “We can do 200 searches in a second, and that means looking over images of at least two million people. What is more important is that we can scale the solution to one billion people or two billion people and the search time will still be less than one second,” he says.
Most facial recognition software also finds faces within the same amount of time. Like an Olympic race, it is often the difference of a second or two between competing firms, with accuracy rates ranging from 70% to over 90%.
NIST routinely evaluates facial recognition technologies with a goal of creating standards within the industry. For IARPA, which has a different mandate than NIST, the goal is to find the best facial recognition solutions for American intelligence agencies. In a very real sense, NTechLab’s best-in-class algorithm is a way for IARPA and the 16 intelligence agencies under DNI to imitate it for its own use. Russia’s private sector could be seen helping to improve U.S. spyware.
IARPA is also developing its own facial recognition technologies. Boehnen’s Janus program has grants out to University of Maryland; University of Southern California; and a Woburn, Mass. based company called Systems & Technology Research. Systems engineers there have been using photos and videos of public figures pulled from the internet for their research. The four-year program ends next September.
“We are just a group of scientists at IARPA,” says Boehnen. “Who wins and who loses is not a political decision.”
Russia has been lambasted by Washington ever since the election of Donald Trump. Russian spearphishing programs compromised a number of computers, including those at the Democratic National Committee, leading to the leak of candidate Hillary Clinton’s emails, according to one recent review by Dell’s SecureWorks. The release of those emails is believed to have cost her the election. It is unclear whether these computer system breaches were ordered by the Kremlin.
Trump’s campaign team is under investigation to see if any members colluded with the Russian government to paint Hillary in a negative light. There have been no serious allegations made against Trump, but his former campaign chairman, Paul Mannafort, was indicted on separate issues related to Ukraine. Trump’s ex-National Security Advisor Mike Flynn may face a similar fate, though also unlikely to be directly linked to Russia or the campaign, said a source with knowledge of the investigation who could not speak on the record.
NTechLab’s win is good for Russia. It at least shows that Russia is indeed more than one big gas station, as Senator John McCain likes to call it. IARPA’s recognition also highlights the role Russians play as influential innovators in the global tech community.
“Some companies still don’t like to work with us bad Russian guys,” Kuharenko says, facetiously. “Sometimes all this bias and opinion about Russians is just because of politics, but we are still working with Americans. But still, there are companies that won’t work with us because we are Russians,” he admits. “I’m not very happy about that. It would be much better if we got along.”
The IARPA contest is a reminder of the importance of a global talent pool to maintain America’s tech prowess. High tech firms and specialty research institutions often count on highly skilled foreign engineers to complement their domestic team. While there is no indication that such visas for high skilled foreign workers in science and technology are on the wane, Silicon Valley has continued to lobby in favor of the more notorious H1-B visa, a visa used overwhelmingly by Indian IT outsourcers like Infosys at this time.
When jokingly asked if he wanted an H1-B visa to come to the U.S. and develop an algorithm to assure Hillary Clinton doesn’t run for office again, Kuharenko chuckled and played along. “I think there’s a different Russian company that does that,” he says.