Researchers from the University of Washington and the University of California, San Diego, wanted to find out if smart cars can recognize individual drivers based solely on their driving habits. Based on an initial test, the researchers announced that existing smart car technology can identify drivers with 100% accuracy.
The research team published their findings on automobile driver fingerprinting in the most recent issue of Proceedings on Privacy Enhancing Technology.
The four scientists recruited 15 drivers for the experiment. First, each person drove a car around a parking lot. Then, each driver navigated a 50-mile loop of Seattle, WA. The team then collected the data from the car’s CAN bus, the in-vehicle computer that manages the digital functions of modern vehicles.
Whether they realize it or not, most drivers have a hidden USB port underneath their steering wheel, which can be accessed by mechanics for repairs and software debugging. They collected data from 15 different sensors in the car, then used a machine learning program to analyze the data.
The researchers’ algorithm was able to identify drivers based on their unique style of driving with 100% accuracy. Not only that, but the algorithm could still identify drivers with nothing but the brake sensor data.
Although the sample size in the experiment was small, the results suggest that different drivers have such unique braking styles that this information can easily be used as a driver fingerprint. After the brake sensor, data from the maximum engine torque sensor was the most useful.
According to the paper, “This signal is the calculated maximum torque that the engine can provide under the current circumstances (altitude, temperature, etc.), based on wide-open throttle conditions.”
Torque sensors are commonly used in all manner of machines, from automobiles to robots. They can withstand extreme conditions, including temperatures ranging from -452 degrees Fahrenheit to 450 degrees Fahrenheit. After the torque sensor, steering wheel data was the most useful.
The research suggests that as the auto industry moves towards smart cars, a balancing act is required between safety and privacy. Already, Toyota has said it will deliver smart cars with artificial intelligence by 2020.