What You Should Know About Face Recognition Technology Used by Police and Spy Agencies Like the NSA
June 04, 2014
The NSA's spying doesn't stop at texts, emails, phone metadata, World of Warcraft, eavesdropping on foreign officials or archiving every phone call made by everyone in the Bahamas. The spy agency is also collecting millions of images every day to plug into face recognition technology, according to a New York Times report based on documents obtained by Edward Snowden.
The NYT notes that the agency has grown increasingly reliant on face recognition technology in the past four years, capitalizing on the vast amount of images passed around through text, email, even video conferencing.
Although many government agencies, including the FBI, DHS and some local police departments use face recognition programs, the NSA has the added advantage of holding a colossal amount of extra information that can be linked to the image, the report notes. A Powerpoint presentation cited in the article showed images of several men along with their known associates and what informants have said about them.
An NSA spokeswoman told the Times that collecting images of Americans through their communications would need court approval (unless they're communicating with someone abroad). But the report points out that it's not clear whether images of Americans are being archived in databases, since there are no rules governing the collection of facial images by any government agency. The spokeswoman declined to say whether the agency grabs images of Americans from social networking sites.
AlterNet spoke with surveillance and face recognition expert Jennifer Lynch of the Electronic Frontier Foundation about the privacy dangers posed by the technology.
Tana Ganeva: How does face recognition work and how is it generally used?
Jennifer Lynch: Face recognition is basically like a mathematical formula or algorithm that's applied to the face. So you start with a photograph of somebody, then apply an algorithm that measures things like the distance between the eyes, and the size of the nose, and how far away the ears are and that sort of thing. And then that mathematical algorithm is stored and compared against images that are later inputted into the system.
So there's two different ways to use facial recognition. One is verification system, where you start with a database of people you've already approved into a system and then somebody presents themselves and says I am Jennifer Lynch and the system looks at the photograph of Jennifer Lynch and says yes, this person is the same, or not the same. That's a verification system that's generally used for gaining entry to a building or maybe using an ATM card.
The second way it's used is as an identification system. And that's where you don't necessarily know who the person is. You have a database of images, face recognition data, and you take a picture of somebody and check them against the database to see who that person is. So the face recognition systems that are in use by the federal government and criminal justice agencies are identification systems and they use large databases and try to identify mugshot photos in general and sometimes other photographs by matching them against the database of facial recognition images.
TG: So, it's not all that surprising that the NSA is collecting images to search with facial recognition because, as you mentioned, so many other law enforcement and government agencies use the technology. Why is face recognition so useful to surveillance and law enforcement agencies?
JL: One reason is that there are so many photographs of people out there in society. We communicate through photos and videos so much more frequently now. It seems like based on the New York Times article and the materials that Snowden has released, that that's one of the reasons the NSA wants to use facial recognition, to augment some of the investigations that they're already doing. And I think another reason is that law enfacement has this belief that it should have as many ways as possible to identify somebody.
The federal government has been using fingerprint databases for well over 100 years. And fingerprints are relatively accurate at identifying people but I think there's this belief that if you have access to other types of technology that your system of identification will be more accurate.
TG: What are the particular advantages of face recognition for law enforcement?
JL: I think the biggest advantage from a law enforcement perspective is that you can take a photograph of somebody from a distance without their knowledge and identify them. At this point facial recognition technology is not at a point where we can just identify everybody in a crowd of people. Even if we had a mugshot photograph of everybody in the crowd of people, the system works best if the lighting conditions are controlled, the angle of view is controlled. But ultimately that's the goal of law enforcement, to be able to identify people as they move through society.
TG:What are the dangers of that?
JL: The biggest danger of that is implications for First Amendment protected rights. We have a strong belief in the United States that you should be able to participate in society anonymously. We come from a history where the founding fathers wrote letters and papers anonymously and that was a very important part of being able to engage in political debate. If we have a society in which surveillance cameras are focused on us all the time and can identify us, then people will be less likely to engage in public debate or participate in public protest, or to associate with people who they may not know or might be different from themselves. And it creates a chilling effect on speech.
TG: In the NYT piece, the NSA spokeswoman said they'd need a court order to collect images of Americans, from their communications at least. But based on how this technology works, what are the chances Americans aren't being sucked into it?
JL: I don't know how it could be possible to ensure that American images are not collected by the NSA. They are just doing complete blanket surveillance. And the way their systems are set up to collect communications data, they collect way more data than they need and I'm sure that includes face recognition data.
TG:The spokewoman seemed pretty slippery on this issue.
JL: Yeah they were very slippery about what they have. It sounded like the spokesperson said concretely that they don't access drivers license databases but the largest face recognition database in the United States is the State Department and it would be very difficult to .... I have a hard time believing that the NSA doesn't have access to the state department's face recognition database.
TG: What's the relationship between the government and the private sector in the development of the technology?
JL: The government is funding a lot of the advanced development of the technology through the private sector. So for example the FBI has been building out its Next Generation Identification biometric database and it has been working with—the contract is with Lockheed Martin but the subcontract is to an organization called Morpho Trust in the United States and Morpho is responsible for a majority of state DMV face recognition databases. And its parent company is a French company and that company has been building out face recognition databases for governments all over the world. So the funding goes into these private companies and they develop solutions for governments and they also use technology funded by the federal government to create products that they then sell to other commercial organizations.
TG: How do you envision the technology advancing in the near future?
JL: I think what we've seen in research and development is that universities and private companies are working on some of the problems of facial recognition right now, so being able to identify people from any angle or very different lighting conditions, being able to identify people as they age, or from photos when they are younger. Identify different minority groups. The same databases research institutions have been using have been primarily Caucasian individuals. Now there's a big movement to improve systems so they can identify minority groups and different ethnicities.
TG: Where's the law stand on this? Are there realistic efforts to bring in some regulation?
JL: There have been some efforts at regulation but I think that Congress is not at a point where it can pass any legislation let alone legislation that would place limits on law enforcement's ability to collect facial recognition data, so I'm not sure how much progress we'll see on that front. We just haven't seen any court cases address facial recognition. So unfortunately what we see with face recognition technology is what we see with a lot of law enforcement surveillance technology which is that police departments go out and purchase this technology and use it before the law has a chance to catch up.
