A team of academics says that you have to trick moderm smart car drive on auto-pilot mode, and those undesirable actions, such as braking or a turn in the direction of — and only through the projection of 2D-images of the road, on the road surface, or by the inclusion of some “triggers” in the video are the billboards.
In this new study, published last week, examined how the company’s advanced driver assistance systems (ADASs), and the automatic pilot technologies, to respond to any 2D objects (also known as a depthless objects, or phantoms) are to be projected into the path of a moving car, and the auto-pilot sensors.
For their study, titled “The Phantom of the ADAS: the Phantom Attack, on the Driving Assistance Systems“[[[[PDF]the researchers have studied only two of the technologies — the Mobileye 630 PRO (used to have it in the Honda, Mazda, Renault, cars, etc.), and the Tesla Model X, HW-2.5 autopilot system.
The researchers found that there is a number of scenarios in which such systems would recognize the 2D projections, such as traffic-related counters, as opposed to what they are projections.
For example, a car would make it possible to identify the 2D-image of human beings projected on the ground, and the pilot was going to be on the break to slow down or stop the vehicle.
The same trick would work with the “projection” of a car.
In the second case, the research team projected the new line on the way, the trick is the car needs to be set to match the lanes of traffic, a trick that would be able to make cars to enter the area of lanes and the location of the client’s life is in danger.
But the trick worked and the projection of 2D objects on the road. For example, for the projection of images from traffic signs to road surfaces, such as walls, trees, and worked just as well, and it has been tested autopilot systems couldn’t tell the difference between the forecast and the actual traffic signs.
In addition, some of the 2D objects, but you don’t always have to be projected. The research team said that the integration of a flat 2D object in a different medium and it works just fine, such as a video advertisement on a billboard. In their experiments, they can be embedded in a road sign, a billboard advert, love the car, it was in a higher speed zone to the next.
The researchers say that the objects that can be projected on a screen in a variety of ways, with the help of a budget of $300 projectors that you can buy on Amazon. These projectors can be handheld, or installed in, on flying drones.
In addition, the research team says that the project of a rogue, and a 2D objects, does not necessarily mean that the projections will need to be displayed for long periods of time. A few hundred of milliseconds is enough, ” she said.
A short burst of the projections, it would be invisible to the human eye, but they could still be seen and is picked up by the high-performance sensors, and video cameras are being used by ADAS and autopilot systems.
This will open the door to real-world scenarios, where the drivers would not even be in the top of the projections, however, the car would suddenly break down, or send you in the direction of the oncoming traffic.
This is an important finding, because the majority of car manufacturers recommend that drivers use the autopilots and only under direct supervision. Car salesmen say that the systems have to be used to help drivers during the driving, but the hands on the wheel and eyes on the road. Since the projections are able to be flashed at the front of the car in front of a few milliseconds, so that drivers who do not comply with these rules, you will not be able to react in time.
The research team also said that it notified both Mobileye and Tesla, on their findings, but the authors have to make improvements to their systems. One of the reasons for this is that there is no way that the software is able to detect the difference between a board that is projected on to a tree, the leaves, and it is a legitimate traffic sign (s) taped to a tree (or a wall), where some of the signs are expected to be found. The same is true for the difference of the projections of the markings of a legitimate way of lines.
However, the autopilot systems will need to be able to distinguish between the projections of the people and the vehicles from the real thing. In this case, the means of locomotion, that is, the auto-pilot and advanced driver assistance systems to support the recognition of 2D, depthless objects, similar to the way in which some facial recognition systems may detect the image of a human face, and to prevent the occurrence of a successful authentication.