We know where you’ve been: Ars acquires 4.6M license plate scans from the cops
Photo by Josh Kellogg
OAKLAND, Calif.—If you have driven in Oakland any time in the last few years, chances are good that the cops know where you’ve been, thanks to their 33 automated license plate readers (LPRs).
 Now Ars knows too.
In response to a public records request, we obtained the entire LPR dataset of the Oakland Police Department (OPD), including more than 4.6 million reads of over 1.1 million unique plates between December 23, 2010 and May 31, 2014. The dataset is likely one of the largest ever publicly released in the United States—perhaps in the world.
After analyzing this data with a custom-built visualization tool, Ars can definitively demonstrate the data’s revelatory potential. Anyone in possession of enough data can often—but not always—make educated guesses about a target’s home or workplace, particularly when someone’s movements are consistent (as with a regular commute).
For instance, during a meeting with an Oakland city council member, Ars was able to accurately guess the block where the council member lives after less than a minute of research using his license plate data. Similarly, while “working” at an Oakland bar mere blocks from Oakland police headquarters, we ran a plate from a car parked in the bar’s driveway through our tool. The plate had been read 48 times over two years in two small clusters: one near the bar and a much larger cluster 24 blocks north in a residential area—likely the driver’s home.
“Where someone goes can reveal a great deal about how he chooses to live his life,” Catherine Crump, a law professor at the University of California, Berkeley, told Ars. “Do they park regularly outside theLighthouse Mosque during times of worship? They’re probably Muslim. Can a car be found outside Beer Revolution a great number of times? May be a craft beer enthusiast—although possibly with a drinking problem.”
In August 2014, the American Civil Liberties Union and the Electronic Frontier Foundation lost a lawsuit to compel the Los Angeles Police Department and the Los Angeles Sheriff’s Department to hand over a mere week’s worth of all LPR data. That case is now on appeal.
In Oakland, OPD’s current LPR dataset shows only a few data points for most vehicles. But there are exceptions—such as the car seen 459 times over two years on a certain block of Chabot Road, east of the Rockridge BART station (one of the city’s richer areas). As Oakland and cities like it deploy more LPR cameras, such datasets could quickly grow more complete.
“Project forward to a world where LPR technology is cheap and they can be mounted on every police car and posted at every traffic light,” said Crump. “Do you think that anyone with a badge should be able to search through that data at their discretion? If not, then you should support restrictions on how long law enforcement agents can store this data, and who can access it, and under what circumstances.”
A low “hit rate”
Specialized LPR cameras mounted in fixed locations or on police cars typically scan passing license plates using optical character recognition technology, checking each plate against a “hot list” of stolen or wanted vehicles. The devices can read up to 60 plates per second and typically record the date, time, and GPS location of any plates—hot or not. (There have been incidents where LPR misreads have led to dangerous confrontations.) Some cities have even mounted such cameras at their city borders, monitoring who comes in and out, including the wealthy city of Piedmont, California, which is totally surrounded by Oakland.
LPR collection began in Oakland back in 2006, and an early OPD analysis showed that the overwhelming majority of the data collected was not a “hit.” In April 2008, the OPD reported to the city council that after using just four LPR units for 16 months, it had read 793,273 plates and had 2,012 hits—a “hit rate” of 0.2 percent. In other words, nearly all of the data collected by an LPR system concerns people not currently under suspicion.
Despite this, in that same report, then-OPD Deputy Chief Dave Kozicki (who has since retired) dubbed the LPR setup an “overwhelming success.” Today, OPD’s LPR hit rate has fallen slightly, to just 0.16 percent.
In addition to LPR data, Ars obtained a list of OPD vehicles and found that the most frequently seen one is plate number 1275287, a 2007 Crown Victoria marked patrol car. Between January 15, 2012 and May 31, 2014, the OPD scanned that vehicle 879 times all over town, primarily in the downtown and North Oakland areas. In fact, nearly all of the 100 most frequently seen cars were other OPD vehicles scanned several hundred times each.
Law enforcement policies vary widely as to how long LPR information can be stored. In California, the wealthy Silicon Valley city of Menlo Park (home to Facebook) retains data for just 30 days. By contrast, the Los Angeles Police Department (LAPD) retains data for two years.
Neither the Oakland City Council nor the OPD has ever imposed a formal data retention limit, though OPD has deleted older LPR data as needed to make room for newer data. As LPR devices and storage prices continue to fall, it’s likely the volume and rate of such data collection will continue to rise, and its retention time can become longer.
“If I’m law enforcement, I would keep it forever,” Brian Owsley, a former federal judge turned law professor at Indiana Tech, told Ars. “That’s the privacy advocates’ concern is that this stuff goes into a database—gigabytes are essentially free now—and this stuff stays forever.”
There’s no evidence that the OPD has abused its database. But absent any strict controls, auditing, or even basic guidelines, it’s hard to know what might or might not have been done.
“Anyone can get this information”
To make sense of the LPR data—which was originally provided in 18 separate Excel spreadsheet files with hundreds of thousands of lines each—Ars hired Mike Tahani, a Bay Area data visualization specialist. Tahani created a simple tool allowing us to search any given plate and plot the locations on a map.
 We did not use the data for any purpose beyond our journalistic attempt to understand what such a large license plate reader dataset reveals. While OPD and other law enforcement agencies have the ability to match a given plate with registration records from the Department of Motor Vehicles and the National Crime Information Center, revealing a car’s owner, Ars does not. In cases where we searched a known individual’s plates, we did so only with their explicit consent.
When shown data on their own movements, local residents had widely varying reactions.
“Doesn’t bother me personally,” said Jon Kaufman, an Oakland resident. “I have nothing to hide.”
Howard Matis, a physicist who works at the Lawrence Berkeley National Laboratory, told Ars that he didn’t know that OPD even had LPRs. With his permission, we ran his plate and showed him a map of the five instances where a camera had captured his car, guessing that they were near where he lived or worked. Matis replied by e-mail: “You are correct, they are places that I and my wife go all the time.”
Matis wasn’t worried about OPD capturing such data, but he was less comfortable knowing that the data was released to the media.
“If anyone can get this information, that’s getting into Big Brother,” he told Ars. “If I was trying to look at what my spouse is doing, [I could]. To me, that is something that is kind of scary. Why do they allow people to release this without a law enforcement reason? Searching it or accessing the information should require a warrant.”
Matis immediately fired off an e-mail to Dan Kalb, his city council member:
Dan,
Do you know why Oakland is spying on me and my wife? We haven’t done anything too radical or illegal.
I gave my license plate to a journalist and he found my wife’s and my car in their database. One of the locations is right near our house.
The astounding thing about this information is that anyone, and I mean anyone, can get this information. Some of the information is more than two years old.
I can see lawyers using this information for lawsuits. I can check where my wife is located. Car companies can see my habits. Insurance companies can check up on their clients. We have entered the world of 1984 with the difference that anyone can get the information.
Ars contacted every member of the Oakland City Council, including newly elected mayor Libby Schaaf, to show them the Oakland LPR data. Dan Kalb, the recipient of Matis’ e-mail, was the only council member who agreed to meet. (Neither Mayor Schaaf nor the recently departed mayor, Jean Quan, responded to requests for an interview.) Kalb represents District 1, which includes some of the city’s richer neighborhoods—including Oakland Hills and Rockridge—and other less affluent regions in the city’s northwest.
 Seated in Kalb’s cluttered City Hall office, Ars explained the LPR issue. We asked for Kalb’s plate number and, within seconds, showed him what the OPD knew about his travels. Our tool revealed that OPD had seen the councilman 51 times between May 2012 and May 2014.
On 16 occasions, Kalb’s car was scanned parked on the street just outside City Hall in the spaces reserved for council members. On another 20 occasions, at various times of day, Kalb’s car was spotted in a tight group on a certain block in the Temescal neighborhood. When Ars guessed that this block must be where he lives, he said that it was.
“I knew these things existed, but I had not delved into the level of detail that you’re sharing with me,” he said.
Kalb is relatively new to City Hall, having only been elected in 2012. Though he did know that the city had LPRs, he said he didn’t know the extent of its usage.
“My awareness is that we have something like this, these mobile LPRs, and I presumed that their primary purpose was to track down stolen vehicles or assist in the investigations of other crimes that knowing the license plate would help,” he told Ars. “It raises the question: what’s the purpose of retaining records for a long period of time?”
A window into the past
The purpose, of course, is to enable retroactive investigation of movements.
Anthony Finnell, the head of OPD’s civilian oversight body (known as the Citizens’ Police Review Board), told Ars that this wasn’t a serious privacy concern.
“I can see as a former investigator that there would be a tool, if a crime had happened you may want to put that person’s plate in and backtrack,” Finnell, a veteran of the Indianapolis Police Department, said. “That doesn’t discount a person’s concern. But I can just say from a personal standpoint, I don’t see it as a big problem, because there’s so many other ways to track somebody.”
Finnell and Kalb are each acutely aware of Oakland’s long-entrenched crime problems. Kalb has even been a victim of crime. A month before the November 2012 election, after walking home from an anti-crime meeting, he was held up at gunpoint for his wallet and iPhone.
Oakland has a higher murder rate than Los Angeles. For three years running, the city has held the dubious honor of being America’s urban robbery capital. The city recorded 4,922 robberies in 2013—or 14 per day. Police see LPRs as one more tool used to put a dent in the crime rate.
Anecdotally, cops often say that LPRÂ data is great for finding stolen vehicles or locating criminal suspects, and it’s easy to imagine how such license plate reader data could be correlated with financial information, cellular data, or even surveillance video in a more complex investigation.
One LPR case often cited on the East Coast is the 2007 murder of five people in Fishkill, New York. Local police were able to access the LPR records of a state trooper who happened to be nearby. That data showed the suspect’s car was in the vicinity of the killing when it happened, which dismantled his alibi.
“They were able to coordinate GPS positions and time and place that car in front of the house shortly before the alarm was raised with the local fire department,” Pete Kontos, a New York State Police senior investigator, told the Hudson Valley Journal News in 2009. “It was a substantial piece of evidence used in the trial.”
Oakland police captain Anthony Toribio, a department employee for 25 years, echoed this sentiment. He told Ars that LPRs are useful to his work.
“It’s a significant tool to have because it can speed up the investigative process by identifying vehicles, linking them to crimes, linking them to locations where that car has been flagged so to speak or identified by the LPR system,” he said. When he served as an area commander, he “used it frequently.”
Toribio noted that Oakland purchased 20 new cameras in 2014, and he said that the city was exploring the use of trailer-mounted “fixed cameras” that could be moved as needed.
“We try to strategically place these cars in areas or beats that are high crime areas,” he said. Toribio cited an occasion when, as a result of a high incidence of burglaries in the hills of North Oakland, he once ordered an LPR-equipped car to that area.
“Often times you have the suspects that are casing the area and driving around, and you never know when a cop car may drive by and they’re not doing nothing at that point, but a short while later they may be driving by and it’s good to have that information.”
Any OPD officer can search the department’s LPR database. While the individual officer’s name is logged, no reason for the search has to be entered. Toribio noted that OPD policy requires officers to have a “legitimate law enforcement purpose, such as following up on a criminal investigation” in order to access the database. “Accessing databases for personal use is a policy violation and illegal in some circumstances,” he added.
Evidence for LPR effectiveness remains largely anecdotal. Toribio could not provide any quantitative evidence to suggest that the LPR system made the police more effective overall at catching suspects than before implementation in 2006. Academics studying the issue say that the question remains open.
“The research is so limited at this point that we don’t know a lot of the answers,” Linda Merola, a criminology and law professor at George Mason University, told Ars. Her university has spearheaded many of the academic efforts on LPRs.
“At this point there have been a handful of evaluations of trying to look at what is the impact in an immediate sense: what is the impact of arrests?” she added. “Can we find more stolen vehicles? Even that research is very limited, and it hasn’t shown very much. What is the impact, if any, on saving data? There have been times anecdotally, but we really don’t know if it serves any purpose other than a theoretical purpose.”
While the potential for LPR dragnet surveillance worries some privacy advocates, the police perspective is that public actions are… public.
In 2006, for instance, Gina Bianchi, the deputy commissioner and counsel at the New York State Division of Criminal Justice Services, wrote in a memo to all local law enforcement agencies across the state, “A license plate reader merely accomplishes, more efficiently, the same task that a police officer may accomplish by reading a license plate and manually entering the number into a database. Therefore, it is reasonable to assume that a court would not hold that the use of a license plate reader would constitute a search.”
Toribio put it more plainly.
“You have to remember the data that we get from the LPR is limited to the plate, the photo, and the location,” he said. “Any information that comes in that alerts us that that vehicle was involved in a crime, the officer has to do more homework. But what expectation of privacy do you have when you’re out in public?”
Toribio is referring to a unanimous 1983 Supreme Court decision in United States v. Knotts. That case famously found: “A person traveling in an automobile on public thoroughfares has no reasonable expectation of privacy in his movements.”
But Knotts came before it was possible to build automated systems that track and store the movement of every car passing by over a period of years. More recently, as part of the unanimous 2012 Supreme Court decision on warrantless GPS tracking, United States v. Jones, Justice Samuel Alito referred to the Knotts decision and said it might not apply as “public” surveillance becomes more comprehensive.
Under this approach, relatively short-term monitoring of a person’s movements on public streets accords with expectations of privacy that our society has recognized as reasonable. See Knotts,460 U.S., at 281-282. But the use of longer term GPS monitoring in investigations of most offenses impinges on expectations of privacy. For such offenses, society’s expectation has been that law enforcement agents and others would not—and indeed, in the main, simply could not—secretly monitor and catalogue every single movement of an individual’s car for a very long period.
While the distinctions between LPR camera systems and GPS tracking are notable, huge LPR databases start to move closer toward the GPS model and might one day run into trouble with the courts. For now, the use of license plate readers has been upheld in numerous criminal cases around the country. And police and prosecutors are happy to use them.
“From a police perspective we are looking at this data to help solve crimes and to help individuals who are solving crimes,” Toribio added. “What I did on a regular basis when reviewing crime reports is look for license plates and then use the license plate reader to get a picture of the vehicle, and its location, and provide that information to police officers to help identify individuals.”
“We have nothing to hide”
In Oakland, no one in the city government has told the police to stop collecting LPR data for years at a time—nor to avoid sharing it. Therefore, the police have done both.
“There is a process that needs to be in place to purge data, and I think that is one of the things that the city is working on, is how long do we keep data?” Toribio said. “We need input from the city attorney and the city needs to decide.”
When asked what would be an appropriate retention period, he replied, “I would say five years is a good amount of time.”
But when Ars told him that the public records request didn’t come back with records older than 2011, he said he would check. Hours later, he called back, saying that the OPD deletes data “as necessary” to free up storage space. Toribio then clarified his position, saying that the LPR data should be kept for two years. (That may explain why some of my own records from April 2012 are missing from the dataset.)
Previously, OPD released my own records to me, but many neighboring communities did not. These organizations cited a law enforcement exemption to the state public records law. So why did OPD share so much data with Ars when other cities have been reticent to do so?
“I know that for OPD, transparency is very important, and unless there is a reason not to release information—if it is part of an ongoing active criminal investigation, for example—we have nothing to hide,” Toribio said. “I think it’s important for a law enforcement organization to be transparent, and it goes to being credible and establishing legitimacy in the community.”
UPDATE Wednesday 1:38am CT: Bryce Newell, a doctoral student at the University of Washington, wrote Ars to say that this collection from Oakland, while large, is not the largest ever released: “I have been examining (for research purposes) an ALPR database from Seattle Police Department disclosed years ago with more than 7.3 million scans (it was originally disclosed to the ACLU of Washington, not me). I have a number of other databases from the Seattle Police Department and Minneapolis Police Department that all have millions of scans (and other, smaller, ones from elsewhere).”