Just months before millions of its internal documents were stolen and dumped on the internet, the Tennessee-based surveillance company Perceptics was preparing to pitch New York’s transit authority on how it could help enforce impending“congestion pricing” rules, according to leaked documents reviewed by The Intercept. The pitch, as outlined in the files, went well beyond mere toll enforcement and into profiling New Yorkers’ travel patterns and companions, creating what experts describe as major privacy risks.
Congestion pricing, on the face of it, doesn’t seem like it would present a privacy risk — it’s a traffic policy, after all, not some new NYPD initiative. The plan is to essentially tax the cars that clog Manhattan’s streets and route the proceeds to public transportation, providing both a deterrent against and palliative for traffic. There won’t be any congestion pricing toll booths: The fee will be assessed automatically and electronically, potentially by photographing the license plates of passing cars and sending the plate owner a bill in the mail. This requires cameras running around the clock, dutifully recording every car that comes and goes. And this, Perceptics claims, is where the company truly shines.
According to an internal presentation released by the Perceptics hacker and reviewed by The Intercept, the company pitched New York’s Metropolitan Transportation Authority, or MTA, in February of this year on how Perceptics’ car-scanning camera arrays, already deployed and honed in areas like the Mexican border and an assortment of U.S. military installations, could help the MTA track down drivers. It’s unknown how the plan was received by the MTA, which administers public transit, bridges, and tolls for New York City and some of its surrounding suburbs, but leaked Perceptics emails show that the company shipped camera hardware to the MTA’s Bridges and Tunnels division for a live demonstration.
Perceptics did not respond to a request for comment. An MTA spokesperson told The Intercept that “all details are still to be determined” regarding congestion pricing enforcement.
The presentation document, titled “Smart Imaging Solutions for New York City Congestion Pricing,” makes clear that Perceptics wants to “produce vehicle-specific profiles” using cameras and “unique machine learning algorithms,” allowing the city to immediately recognize and build travel histories of every car in the congestion zone. Law enforcement and surveillance experts said the system described goes far beyond what would ever be necessary to mail scofflaws traffic tickets. Instead, it is an entirely new sort of surveillance apparatus that tracks deeply personal information like “customer travel patterns and travel consistency,” the number of passengers in the car, or “likely trip purpose,” and associates this information with a unique fingerprint of every vehicle that passes by Perceptics’ cameras.
Allie Bohm, a policy counsel with the New York Civil Liberties Union, described the Perceptics plan as an “incredibly privacy-invasive proposal” that “raises all sorts of associational and First Amendment concerns.” Bohm expressed particular alarm about the possibility of a congestion pricing enforcement system eventually feeding data into the NYPD’s existing surveillance regime. “The NYPD has fancied itself an intelligence agency for a very long time,” said Bohm. “These are folks who are pioneering some really, at best, questionable, and, at worst, alarming programs of surveillance and of drawing conclusions from innocuous behavior.”
The Electronic Frontier Foundation, a San Francisco-based digital privacy nonprofit, has described the technology as “a form of mass surveillance .” Now, a new generation of tech firms has made it possible for private citizens to use the devices, known as automatic license plate readers, or ALPRs—without the strict oversight that governs this type of data collection by law enforcement.
The MTA will not deploy congestion pricing before 2021 and has yet to select a tolling vendor. But whether Perceptics wins a contract or not, its idea to bring to the heart of Manhattan military-grade surveillance technology — already provided to Saudi Special Forces and the Jordanian army, according to a Perceptics document — is an example of how something as innocuous-sounding as congestion pricing can turn into a surveillance sprawl.
The Perceptics proposal works like this: Cars passing through Manhattan’s congestion pricing enforcement zone would be assigned a “unique vehicle identifier,” or UVID, which identifies the car even when cameras can’t read its license plate. “Just as a forensic analyst or database can identify an individual from a partial fingerprint,” explains the Perceptics website, “the UVID can also identify a plate, even when portions of its signature are missing or obscured.” As opposed to typical automatic license plate readers, which only attempt to recognize the letters, numbers, and state stamped on a given plate, Perceptics says its method “sees each image as a visual map of the region of interest,” an amalgamation of the car’s “unique attributes,” as a different document puts it. Perceptics’ holistic approach to vehicle surveillance “can be used to correlate all the data associated with the vehicle, cargo, and passengers as the vehicle travels through the city,” according to the presentation prepared for the MTA. This data is used to create an invisible computerized signature for your car that will be used to track it from that point onward, without even needing the license plate. You’ve heard of facial recognition; this is car recognition.
It’s unclear what exactly “all the data” means in this case, but the MTA pitch shows the use of a Perceptics technology named Vehicle Occupancy Imaging System, or VOIS, which photographs drivers and their passengers in addition to the car’s license plate. In a separate pitch document, Perceptics notes that VOIS is “capable of observing driver behavior, cell phone usage, and seat belt enforcement.” In other words, for situations where Perceptics’ cameras and software can’t read your plate — let’s say there’s a trailer hitch in the way, or an inconvenient shadow — the system can fall back on scanning people in your car, where it’s headed, how fast it was going, and other “unique attributes” of you and your ride. Although the MTA presentation makes no mention of biometrics, other Perceptics documents do; a 2018 “strategic summary” paper notes the “opportunity to broaden Border products with stronger driver camera to support facial recognition biometrics with drivers and passengers,” and internal emails indicate the company had begun using Amazon’s controversial Rekognition technology as of August 2018.
At a time when China and the United States are locked in a rivalry on several fronts including trade and technology, Hikvision – which is the world’s largest surveillance technology company and based in Hangzhou in eastern China – has supplied the equipment and software used by an American force that polices a population of about 8.6 million people.
New Yorkers are already under pervasive surveillance as it stands. A May blog post by Ángel Díaz of NYU’s Brennan Center for Justice noted: “Mass surveillance is not the fictional dystopia of Orwell or only found in China — it is today’s New York City,” a place where “clusters of cameras and sensors blanket nearly every city block, and New York Police Department vans with 3-story surveillance cranes tower above many minority communities across the five boroughs. … The NYPD’s Domain Awareness System alone links with at least 9,000 cameras and 500 license plate readers.” It’s an urban dragnet that’s increasingly impossible to avoid without staying inside your home.
The system envisioned by Perceptics would a layer a new, troubling system on top of those old ones. After reviewing the company’s documents, Díaz cautioned that the proposal represents “the kind of profiling that could give you a pretty detailed look at where people live, where they work, where they pray, [and] where they go to school,” adding that it was “ironic [the technology] was developed for border security,” as the “information would be very attractive to federal immigration authorities.” Díaz further warned that even if this tool is not initially deployed for law enforcement purposes, the datasets can eventually make their way into the hands of police departments: “Once you start collecting the data, even if it’s for one purpose, once law enforcement gets a sense what’s being stored, it can be alluring to ask” to get their hands on it.