Image credit: Peter Blanchard via flickr | http://bit.ly/1q1sNlt. Rights information: http://bit.ly/c34Awz
By: Benjamin Plackett, Inside Science
(Inside Science) -- It may be unintentional, but it’s possible that your car insurer may know a lot more about where you go than you’d like.
A computer scientist at Rutgers University in New Brunswick, New Jersey, has developed an algorithm, which demonstrates that the information gathered by the boxes insurers place in your car in exchange for the possibility of lower premiums is enough to tell them where you've been and where you're headed.
Privacy experts agree this sort of data is likely to have a high economic value if it were sold to a third party. You’ve just arrived at the grocery store? Maybe the cafe next door would like to text you about their lunch specials. This sort of mobile advertising is expected to be worth $20 billion by 2016. That’s roughly the gross domestic product of Cyprus. An insurance industry spokesperson told Inside Science that it is technically possible for them to track customers with data gathered from the speedometer, but they’re not currently doing so.
The Rutgers team's computer program uses a technique called “elastic pathing,” which assumes the drivers' starting point is their own home. It then anticipates their route by using speed data to guess whether they turn left or right at each junction. For example, suppose at the end of your driveway you have two options: turn right onto the highway or left to meander through downtown’s traffic lights. It’s pretty easy to assume that you’re on the highway if you’re cruising along at 60 mph.
“Our algorithm allows us to predict a driver’s destination by comparing the starting point and speed data with a map,” said Janne Lindqvist, a computer scientist at Rutgers.
Lindqvist’s accuracy rate, for the time being at least, leaves something to be desired. His algorithm’s predictions are only correct 26 percent of the time and within one third of a mile. Nevertheless, he said it’s still an impressive result.
“It means a lot because the list of paths a driver can take explodes after the first junction. Within one mile from [the average] grocery story there are more than 1 million potential routes that they could have taken,” said Lindqvist.
There are a few relatively straightforward modifications to be made that would significantly improve the success rate, said Lindqvist. He’d like to include information from Google Maps that monitors traffic jams in real time as well as historical data from a driver’s backlog of journeys. In fact, the way Google knows whether to assign the color red, yellow or green to roads on Google Maps is by tracing the movement and speed of Android cellphones along the route. So if you’re an Android user, Lindqvist’s algorithm doesn’t present any new privacy concerns.
Progressive offers an opt-in program called “Snapshot” to their auto insurance customers. A device collects speed readings to determine mileage and the frequency of sudden stops. They use this information to decide whether you’re a safe driver or not and offer a discount on auto insurance if they like the way you drive.
“We do not use the information in any other way,” confirmed Progressive spokesperson Amanda Lupica via email.
However, Progressive isn’t opposed to collecting and using location data in principle, it’s just that Lindqvist’s method isn’t accurate enough for them at the moment, said Lupica. Progressive is currently toying with the introduction of GPS to their Snapshot system.
“We’re testing hypotheses like whether highway driving is safer than busy city street driving,” explained Lupica. “If this location data proves valuable, it will help us price insurance even more accurately.”
This worries privacy expert and professor of mathematics at the University of Texas at Austin Andrew Blumberg. “It’s ridiculous that you should have to spend more on insurance based on where you drive. Better not go the Bronx one day — otherwise your rate might go up,” he said.
The opt-out clause of GPS tracking and Progressive’s statement that they’re not using speed data to keep tabs on you isn’t the point, said Blumberg. It means we now have to place an undue amount of faith in an insurer that tracks such data, entrusting them to handle sensitive location data properly, he added.
“If I work at Allstate, can I just look up where my girlfriend is?”
Lindqvist is of the same mind.
“We’ve shown that when insurance companies say they just collect speed data, it isn’t sufficient to protect privacy.”
Blumberg would solve the problem with more stringent privacy regulations at a federal level.
“There should be more meaningful penalties if they mishandle our location data,” he said. “Right now there are just minor fines.”
Lindqvist, on the other, hand thinks the solution lies in the data collection itself.
“They could reassess what data they collect. They don’t need speed data to deduce driver behavior. Just data on hard breaks would be enough.”
Benjamin Plackett is a science journalist based in New York City.
Reprinted with permission from Inside Science, an editorially independent news product of the American Institute of Physics, a nonprofit organization dedicated to advancing, promoting and serving the physical sciences.