In a crowded conference room earlier this month in Menlo Park, California, representatives from companies around the world listened intently as officials from the Department of Homeland Security explained the bidding process for contracts to develop facial recognition capabilities at land border crossings. The companies were eager to get on the ground floor of the government’s pilot program for using facial scanners and databases at the border. The pilot program, managed by the DHS Silicon Valley Innovation Program, in conjunction with U.S. Customs and Border Protection, is one of several initiatives to use image-recognition technology designed for security purposes on a grander scale.
For the companies looking to get in on the action, the government has a test in mind: According to solicitation documents, the pilot program challenges bidding companies to accurately identify three individuals in a car traveling at roughly 25 mph through raised car windows and light rain, with one of the passengers in the vehicle’s backseat. While DHS officials acknowledged in Menlo Park that facial recognition is far from a mature technology and would accept up to a 70 percent error rate for initial projects, the short-term goals of both the government and private sectors are to drastically reduce inaccuracies. CBP officials already believe that technology is far more accurate than humans at identifying people.
“Technology is far better at matching identities than humans alone,” said Dan Tancier, the CBP Field Operations official in charge of the biometric program for entry and exit into the country. “That’s a given.”
The U.S. has not yet arrived at scenarios envisioned in science fiction, such as “Minority Report,” where authorities closely track citizens using ubiquitous biometric scanners, but the government’s entrée into using facial-recognition technology is well underway. The most rapid expansions, of the sort foreshadowed in Menlo Park, will come in the U.S.’s immigration infrastructure. Some pilot programs are already underway — Boston’s Logan Airport has one. The new DHS initiative, however, is seeking to bring facial recognition to land borders, posing challenges because, indeed, cars might be driving by in the rain, with all the windows up.
The proliferation of biometric scanning at borders has long been codified: The pilot program featured in Menlo Park is authorized through a 2004 law that encourages the government to invest in biometric security for immigration purposes. But the technology is only beginning to catch up with the government’s ambitions. In turn, with a thoroughly anti-immigrant administration in place, those government ambitions are growing.
The mandate handed down in the 2004 law could soon be just the tip of the iceberg. In January, President Donald Trump signed an executive order calling for the expedited implementation of biometric security procedures for individuals leaving or entering the country. (The order was reissued in March with the same language on biometrics.) As The Intercept reported in January, several facial recognition and biometric lobbyists secured positions advising the administration’s homeland security and immigration agencies.
Experts are raising concerns about the potential for abuses and errors — and especially the preservation of privacy.
Congress may also make sweeping updates to DHS mandates. Border security legislation, sponsored by Rep. Michael McCaul, R-Texas, and Sen. John Cornyn, R-Texas, explicitly calls for border security officials to utilize facial recognition technology “to the greatest extent practicable.”
Experts, however, are raising concerns about the potential for abuses and errors — and especially the preservation of privacy. Harrison Rudolph, a law fellow at Georgetown University’s Center on Privacy and Technology, said the DHS’s facial recognition programs “aren’t ready for takeoff.” Rudolph said that the agency has not elucidated specific rules to limit data collection, retention, and reuse with regard to facial recognition technology.
The lack of privacy policies going forward is all the more urgent because the pilot programs that the DHS already has in place lack such protections. For the program at Logan Airport, the memorandum of understanding between the company and the DHS was kept secret, and no privacy guidelines have been made public. “We’re still waiting on DHS to issue rules to back up its privacy promises,” Rudolph said.
Ari Schuler, an adviser to Customs and Border Protection Commissioner Ronald D. Vitiello, brushed aside those concerns at the Menlo Park conference. The agency, he said, intends to develop the pilot facial recognition program through existing privacy safeguards.
“I can say the department has some very explicit, high-level policies,” Schuler said, responding to an audience question about ethics and privacy. “Of course we have the DHS privacy office which all of this has to go through and be reviewed with respect to impact to privacy.”
Silicon Valley companies like those in attendance at the DHS conference seem unlikely to stand in the way: With an administration bent on immigration enforcement by almost any means, the rapid expansion of government programs to bring biometrics to borders could be a bonanza for technology industries.
With an administration bent on immigration enforcement by almost any means, the rapid expansion of government programs to bring biometrics to borders could be a bonanza for technology industries.
The DHS officials in Menlo Park underscored the need to build better collaborative relationships with the private sector and accelerate the development of surveillance technologies. Schuler, the CBP adviser, told the audience that DHS wants to encourage research and innovation in the biometrics field. “We want competition. We want the industry to do better and better,” Schuler said. “A fragmented industry is very difficult to do business with.” Part of the effort is to engage smaller companies that might not traditionally pursue government contracts, said Melissa Ho, the managing director of the DHS Silicon Valley office, to bring technology “to the field faster.”
Representatives from the industry listened on as DHS officials spoke. Some of the companies in attendance already have experience in public-private endeavors, some even with DHS. An official with JetBlue, which runs the facial recognition pilot program out of Logan Airport, part of a separate DHS initiative, was in Menlo Park. Alex Kaufman, who works for JetBlue’s venture capital arm, expressed interest in the government’s growing investments in biometric technology. That Logan Airport program, Kaufman said, “was received very, very positively.”
The partnerships built upon homeland security, Kaufman noted, could spur a new wave of private-sector innovation. Over the next five years, facial recognition technology “really does enable the next generation of travel in some form or fashion,” Kaufman said. “In terms of the airport as the next generation retail experience, being able to hyper-target your customer, the passenger, who are the retailers, who are the airlines, whoever they might be, the stakeholders, I think that’s pretty exciting.”
The tipping point for facial recognition may be right around the corner as dozens of firms are fine-tuning the technology.
While DHS invites bids on facial recognition technology and pilots multiple test programs at airports around the country, technology firms are cultivating contacts within government. The industry’s outreach was on display at the GTC Technology Conference in Washington, D.C., an event sponsored earlier this month at the Ronald Reagan Building and International Trade Center. Sponsored by chip-manufacturer NVIDIA, the event was billed as an opportunity to introduce the public sector to cutting-edge technology.
Industry representatives mingled with leading lawmakers and military research officials. Rep. Darrell Issa, R-Calif., was scheduled to scheduled to appear, along with officials from In-Q-Tel, the Navy’s Space and Naval Warfare Systems Command research lab, as well as Richard Kidd, the deputy assistant secretary of the Army.
While media attention has focused largely on the new iPhone facial recognition technology and Facebook’s attempts to identify individuals in photographs, advancements in the biometrics field are rapidly reshaping the homeland security industry. At the Washington event, NVIDIA executives boasted that the company’s graphics processors have accelerated the development of machine learning; their work has surpassed current capabilities of facial and image recognition technologies, the executives said. A number of firms that utilize NVIDIA’s chips sponsored panels and exhibits to showcase the technologies’ application for law enforcement and homeland security.
Rapid developments have produced stunning advancements in the field. Recent contests sponsored by the Intelligence Advanced Research Projects Activity initiative and the U.S. National Institute of Standards and Technology have shown that a number of global firms have achieved incredibly accurate facial recognition technology platforms. NTechLab, a Russian firm, was able to match a person’s face correctly 99.9 percent of the time during in a NIST-sponsored contest in October.
Companies at the Reagan Building showed off all manners of surveillance technology that could be useful for the authorities. Coban Technology, which has a history in police technologies, displayed a Delaware State Trooper vehicle and drone on the exhibition floor; the police equipment was outfitted with an array of surveillance equipment designed to provide real-time analysis of potential criminal activity. The firm claims that an advanced artificial intelligence system will allow an array of video feeds, from dashboard cameras to officer body cameras, to instantly identify suspects and locate abducted children, as well as recognize terrorist activity.
— NVIDIA Embedded (@NVIDIAEmbedded) November 2, 2017
Like Coban, many legacy police technology firms are moving quickly to embrace machine learning to offer new services to law enforcement. Axon Network, formerly known as Taser, now advertises the ability to spot criminals and “anticipate criminal activity” using its existing body camera products. This October, the Los Angeles Police Department selected Axon’s Artificial Intelligence Research team to conduct a 14-month trial experiment using machine learning to analyze LAPD’s archive of body-worn and dashboard camera video. The LAPD-Axon collaboration is funded by a Department of Justice grant. (Axon is also making forays into crowdsourced video databases to purportedly solve crimes.) Vigilant Solutions, a pioneer in license plate recognition products used for speed cameras, has begun to market itself as a facial recognition firm.
Along with the legacy companies, several startups at the NVIDIA conference also pitched new efforts to harness the power of security cameras using facial and image recognition technology. Sean Huver, a former DARPA engineer who now serves as chief executive of a company called Deep Science, presented on a product to provide automatic alerts for criminal activity. The technology utilizes existing security camera networks in 7-Elevens and Gap stores to recognize a robbery in progress or other security threats. Huver, during his presentation, noted that his company is currently working on a program to give police direct access to the video feed alerts they are establishing in retail outlets.
Vision Labs, a Russian company that presented at the NVIDIA conference, touted a proprietary facial recognition technology that it said used machine learning to continually adapt and teach itself greater levels of accuracy. “In our database we have at least 12 million unique people,” said Alexander Khanin, the chief executive of Vision Labs. Khanin, whose firm is used by Russian banks and law enforcement to improve security, is currently engaged in effort to pitch American police departments. One slide from his presentation touted “city-wide” surveillance to identify and locate criminal suspects.
Vadim Kilimnichenko, project manager at Vision Labs, told The Intercept that the company does not yet have any law enforcement clients in the U.S. The firm currently collaborates with Facebook and Cisco, and maintains clients such as Equifax and Sberbank.
For now, these companies are looking for pieces of a relatively limited government pie. At the Menlo Park conference, the Silicon Valley firms in attendance were vying for border patrol contracts of up to $800,000. However, officials indicated that the program was only a preview of much larger efforts afoot.
During the Q&A period, Dan Strellis of Rapiscan Systems, an airport security contractor, asked how the Department of Homeland Security will overcome challenges in bringing more stakeholders into the process as it begins to embrace new technological solutions.
“The biometric security is the big spearhead now,” said the CBP’s Schuler, pointing to the agency’s pilot programs. “Once people start making money, having happier passengers, we’ll start to see a gold-rush effect, and there are clearly early adopters who are reaping benefits.”