In addition to radar, cameras or GPS-system,
Google decided to install on the roof of his unmanned vehicle also a powerful eye, able to have a 360 degree view. The aforementioned eye is
LIDAR (Light Detection and Ranging), which is able to measure distances using laser light, as a result of which a 3D map of everything that surrounds the car is formed.
Despite this technology, which allows the car to go on the road without a driver, eliminating any errors peculiar to
people-drivers , however, the manufacturers of such autonomous cars do not celebrate the victory, because
LIDAR sensors are not completely perfect. Jonathan Petit, a security expert, demonstrated their vulnerabilities and showed that they can be easily deceived.
The expert managed to fool the sensor using a
laser pointer and a pulse generator , while he also stated that Raspberry Pi or Arduino could be used instead. So, to fool a smart car, all you have to do is spend about $ 60 (or 53 euros).
')
Thanks to this system,
potential hackers could make the machine believe that there is a wall next to it, a person or another car, as a result of which it will slow down. You can also send false signals and
force the car to stop completely to eliminate the possibility of a collision with these non-existent objects.
While radars operate at their own frequencies, which makes them less vulnerable, Petit was able to easily record and
simulate laser pulses generated by the LIDAR system. He was able to make various copies of false obstacles and even forced them to "move", thereby confusing the sensor and making him believe that this illusion is a reality
at distances from 20 to 350 meters .
Petit will present the details of her research at the upcoming Black Hat Europe conference, which will be held in Amsterdam in November. However, at the moment all that he has demonstrated, shows:
one of the main drawbacks of these machines is their vulnerability.
Google unmanned vehicles use LIDAR technology, which was developed by
Velodyne , located in Silicon Valley. She developed a device capable of storing
more than a million pieces of information per second , which allows the car to continue its movement without incident.
Although this invention is not cheap. Each device costs 85,000 dollars (about 75,000 euros). At the same time, this study shows that a high price does not always mean a high level of protection and security: even the most expensive of them are at risk.
Although attacks are limited to a specific device at the present time, the expert claims that all manufacturers should pay special attention to safety issues and take the necessary measures to avoid accidents on the roads. “If a self-driving car has bad input data, it will lead to management errors,” says Petit.
The problem can be solved with a better detection system: “A quality system can cross-check each detection with other data, and then filter out improbable detections. But I do not think that machine manufacturers have already done that. So it could be a good bell for them. ”
And not only
Google tested these LIDAR systems: Mercedes, Lexus and Audi also tried their prototypes on their cars, but because they also need to pay attention to any possible security risks, if they want their unmanned vehicles to be the next step in the development of the automotive industry .